원클릭으로 Manus에서 모든 스킬 실행

project-discovery-workflow

스타4

포크0

업데이트2026년 6월 7일 05:16

ProjectDiscovery toolchain operator reference. 20 CLI tools (pdtm tool manager, subfinder, naabu, katana, cloudlist, chaos, uncover, alterx, httpx, dnsx, tlsx, nuclei, interactsh, cvemap, notify, mapcidr, cdncheck, proxify, simplehttpserver, aix) organized as a 4-phase pipeline (Discover, Enrich, Detect, Notify) chained via stdin/stdout. Covers current flags, footguns, provider-config.yaml auth (PDCP_API_KEY, SHODAN_API_KEY, CENSYS_API_ID/SECRET, FOFA_EMAIL/KEY, QUAKE_TOKEN, HUNTER_API_KEY, ZOOMEYE_API_KEY), Nuclei v3 changes (-jsonl, -duc deprecation, -auth, flow/code protocols, self-contained templates, authenticated scanning via -secret-file), rate-limit tuning matrices, modern chains (SPA crawl with katana -headless, KEV-filtered nuclei via cvemap, JARM cert clustering with tlsx, ASN sweep via uncover, PDCP cloud upload with -pd / -auth), interactsh self-hosting on a VPS (NS records + -domain + -t token), notify provider-config webhooks, and PD-vs-BBOT positioning. Use when wiring up bug-bounty or pentest

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

cmndcntrlcyber

cmndcntrlcyber/rtpi

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

ProjectDiscovery Toolchain — Operator Reference

ProjectDiscovery (PD) ships 20-some single-purpose Go CLIs that follow the Unix philosophy: each does one thing, all chain on stdin/stdout, and all agree on the same handful of conventions (-list, -silent, -json, -rl, -config, ~/.config/<tool>/). The result is a recon pipeline you can compose with | and tee, rather than a monolith.

This reference is grouped into the four phases the tools fall into — Discover → Enrich → Detect → Notify — plus a tool-manager (pdtm), a Pipeline Cookbook, a cross-cutting Footguns table, a centralized Provider Auth table, and Modern Patterns covering Nuclei v3+ flow, the ProjectDiscovery Cloud Platform (PDCP), and PD-vs-BBOT positioning.

When to reach for PD vs a framework like BBOT: PD wins when you want a transparent pipeline you read top-to-bottom in your shell history — small Unix tools you can swap in and out, with output you can grep and diff. BBOT wins when you want one tool that orchestrates a hundred modules and produces a unified asset graph. Most operators end up running both — BBOT for hands-off attack-surface monitoring, PD tools when actively hunting.

pdtm — Tool Manager

Use pdtm for everything. Skipping it and go install-ing each tool individually is the most common source of "I have nuclei v2, why don't templates resolve?" pain.

go install -v github.com/projectdiscovery/pdtm/cmd/pdtm@latest  # needs Go 1.24.3+
pdtm -ia          # install all PD tools
pdtm -ua          # update all
pdtm -i nuclei,httpx,katana   # install a subset
pdtm -sp          # show the binary path
pdtm -up          # self-update pdtm itself

Binaries land in $HOME/.pdtm/go/bin, which pdtm appends to $PATH on first run. Restart the shell after install — that's the "I tried to use httpx and it won't work!" footgun. Pin a tool via go install …@v0.0.X only when you specifically need to reproduce a release; otherwise let pdtm -ua drive.

Phase 1 — Discover (map attack surface)

subfinder — passive subdomain enumeration

subfinder -d target.com -all -recursive -silent | tee subs.txt

Key flags (full list via -h):

-d / -dL — single domain or file of domains.
-all — uses every configured source (slow, more dupes; default uses a curated subset).
-recursive — only sources that can recurse on found subdomains.
-active — DNS-resolve discovered names; required for -oI (include IP) and -nW (filter wildcards).
-pc, -provider-config — point at a different provider-config.yaml.
-s, -sources / -es, -exclude-sources — narrow or block sources; subfinder -ls lists the ~30 supported.
-rl — HTTP requests per second across sources.
-t — concurrent resolution goroutines (only used with -active).
-oJ, -json — JSONL output. Combine with -cs to include source attribution per record.
-r / -rL — custom resolvers (file or comma list).

Footguns:

-all without API keys returns a lot of junk from the noisier free sources. Configure providers first, then -all is actually useful.
Default resolvers (8.8.8.8 etc.) get rate-limited fast on big sweeps. Use your own resolver list (trickest/resolvers, proabiral/Fresh-Resolvers) and update it weekly.
DNS caching at the source side can hide newly added subdomains — passive enum is a leading indicator, not a real-time signal.
subfinder -d example.com -oI -active returns nothing if -active isn't paired with -nW; the IP-include path only fires on resolved records (see GitHub discussion #1126).

Provider auth (~/.config/subfinder/provider-config.yaml):

Free-tier keys worth getting: chaos, virustotal, securitytrails, censys (250/mo), github, fullhunt, intelx, leakix, netlas.
FOFA wants email:key, censys wants api_id:api_secret, github wants raw PATs.
passivetotal uses email:password.
Multiple keys per source rotate automatically.

naabu — fast port scanner

sudo naabu -host target.com -top-ports 1000 -exclude-cdn -nmap-cli 'nmap -sV -oX scan.xml'

Key flags:

-host / -list / -eh / -ef — input, exclusions.
-p, -tp top-ports {100|1000|full}, -ep, -pf — port selection.
-s SYN|CONNECT — scan type. SYN needs root; otherwise it silently falls back to CONNECT.
-c 25 — worker threads; -rate 1000 — packets/sec.
-sV / -sV-fast / -sV-workers / -sV-probes — built-in service version detection via the nmap-service-probes file.
-ec, -exclude-cdn — for Cloudflare / Akamai / Incapsula / Sucuri IPs, only scan 80/443.
-cdn — annotate detected CDN in output.
-passive — use Shodan InternetDB (no packets sent).
-nmap-cli 'nmap -sV …' — pipe found ports into a real nmap.
-metrics-port — exposes JSON scan stats on localhost.
-iv 4,6 -sa — scan all v4+v6 IPs a host resolves to.

Footguns:

Default config assumes a VPS — local boxes need lower rates or you saturate your uplink and the scan looks "slow" while you're actually ddosing yourself.
-exclude-cdn is required for Cloudflare/Akamai etc. unless you want to fingerprint the CDN, not the origin. Without it, you scan the CDN's IP pool and waste a lot of packets for cloudflare:443 banner outputs.
Without sudo you get CONNECT scans, which are slower and noisier (full TCP handshake, more IDS hits).
Masscan goes faster but is less reliable; naabu is the right default for bug-bounty-scale lists where you re-probe with httpx anyway.

katana — next-gen crawler

katana -u https://target.com -d 5 -jc -ps -iqp -kf all -c 30 -ef css,woff,svg

Key flags:

-u / -list — input.
-d 5 — crawl depth.
-c 30 -p 10 -rl 100 -rlm 500 — concurrency, parallelism, rate limits (req/s and req/min).
-jc, -js-crawl — parse endpoints out of JS bundles (the killer feature on SPAs).
-hl, -headless — Chromium-backed crawl for React/Vue/Angular. Combine with -sc, -system-chrome to reuse the host Chrome, -ho, -headless-options for flags, -nos, -no-sandbox in containers, -page-load-strategy {normal|eager|none} to control the wait model.
-cs / -cos / -fs {dn|rdn|fqdn} / -ns / -do — scope control.
-ef css,txt,md extension filter; -ndef disables the default extension filter. -mr / -fr regex match/filter.
-kf all|robotstxt|sitemapxml|… — known-file harvesters.
-iqp, -ignore-query-params — collapses noisy parametric URLs.
-ps, -passive — pull URLs from waybackmachine/commoncrawl/alienvault.
-aff, -automatic-form-fill — fills forms during the crawl (experimental, worth turning on for endpoint discovery).
-jsonl — structured output you can feed back into nuclei/httpx.

Footguns:

Headless mode shells out to Chrome; on big lists you'll spawn hundreds of tabs and OOM. Cap with -p 5 -c 10 first, then scale.
DSL filters apply to jsonl keys, not the URL — -fr against the text URL works, -fr against a key needs the DSL pattern.
katana -u target.com -kf all against SPA-heavy targets without -headless will look "shallow" because most endpoints only emit after JS execution.

cloudlist — multi-cloud asset enumeration

cloudlist -provider aws,gcp,do -exclude-private -json | jq .

Key flags:

-pc, -provider-config — ~/.config/cloudlist/provider-config.yaml.
-p / -id / -s — filter by provider, by provider-config id, by service (e.g. -s ec2,route53,lambda).
-host / -ip / -ep, -exclude-private — output shape and pruning.

Footguns:

Cloudlist is the blue-team tool — it inventories your accounts, not someone else's. Don't put it on a recon checklist for an external target.
AWS supports assume_role_arn + external_id for cross-account scans; if the role isn't trusted to be assumed from your scanning account, the call silently returns zero assets.
GCP has two modes: per-service APIs (fast, project-scoped) and the Cloud Asset Inventory API (org-wide; needs cloudasset.assets.* perms).
Azure needs the explicit minimum role (see PD docs — VMs, NICs, Public IPs, DNS, Storage, AKS, CDN reads).
18 providers supported (aws, gcp, azure, alibaba, digitalocean, scaleway, heroku, linode, fastly, namecheap, cloudflare, hetzner, kubernetes, nomad, consul, terraform, openstack, ibm).

chaos — PD's internet-wide DNS dataset

chaos -d uber.com -silent | httpx -silent -title -tech-detect

Auth: PDCP_API_KEY env var, OR chaos -d uber.com -key XXX. Get the key from cloud.projectdiscovery.io. The old CHAOS_KEY env is gone — use PDCP_API_KEY.

REST: GET https://dns.projectdiscovery.io/dns/{domain}/subdomains with Authorization: API_KEY. Returns {"domain":"…","subdomains":["…"],"count":N}.

Footguns:

Chaos is a snapshot, not a live resolver. Subdomains that resolved hours ago may be stale; subdomains added in the last hour may be missing. Pair with subfinder and your own crt.sh poll.
Bug-bounty programs that publish in-scope hosts to Chaos are the highest-signal targets — chaos -d <program> first, then enrich.

uncover — query Shodan/Censys/FOFA/Quake/Hunter/ZoomEye/Netlas/CriminalIP/PublicWWW/HunterHow

uncover -q 'ssl:"target.com"' -e shodan,censys -ul 500 -uf 'host:port'

Key flags:

-q '…' — query (stdin/file/CLI).
-e shodan,censys,fofa,quake,hunter,zoomeye,netlas,criminalip,publicwww,hunterhow,shodan-idb — engine(s). Default is shodan.
-asq jira — use ProjectDiscovery's curated "awesome search queries".
-uf ip:port|host:port|ip|host — output field shape.
-ul 100 — limit (default 100).
-ur 60 — req/min when an engine doesn't expose its rate limit.
Engine-specific aliases (-s/-shodan, -sd/-shodan-idb, -ff/-fofa, -cs/-censys, -qk/-quake, -ht/-hunter, -ze/-zoomeye, -ne/-netlas, -cl/-criminalip).
Within nuclei: -uc -uq '…' -ue shodan to fan-out templates over uncover results.

Auth — environment variables (or ~/.config/uncover/provider-config.yaml):

SHODAN_API_KEY, CENSYS_API_ID+CENSYS_API_SECRET, FOFA_EMAIL+FOFA_KEY, QUAKE_TOKEN, HUNTER_API_KEY, ZOOMEYE_API_KEY, NETLAS_API_KEY, CRIMINALIP_API_KEY, PUBLICWWW_API_KEY, HUNTERHOW_API_KEY.

Footguns:

Each engine bills queries differently — Shodan free tier is 100 results/mo, Censys is 250/mo. -ul 500 against Shodan eats your monthly budget on a single command.
FOFA requires email:key, not just key; subfinder/uncover both fail silently with "no API key defined" when only the key is set (see subfinder issue #1563).
Shodan-idb is the free InternetDB — no key required, far less data than the paid Shodan engine.

alterx — DSL-based subdomain permutation

chaos -d tesla.com | alterx -enrich -p '{{sub}}-{{word}}.{{suffix}}' | dnsx -silent

Key flags:

-l / -p / -pp — input subdomains, pattern, payload (key=value).
-en, -enrich — auto-extract words from the input as payloads.
-es, -estimate — preview the permutation count before generating.
-limit / -ms 100mb / -o — bound output size.
-ac — point at a custom permutation config.

DSL variables (nuclei-style):

{{sub}} — the input subdomain string.
{{word}} — the payload word.
{{suffix}} — the TLD/eTLD.
{{root}} — the registered domain.

Footguns:

Alterx generates candidates, not results — pipe to dnsx -silent to validate. Without -enrich, you'll generate millions of permutations that mostly resolve to nothing.
The default permutation file lives at ~/.config/alterx/permutation_v0.0.1.yaml; if you've edited it and pdtm later overwrites it, alterx silently picks up the new file.

Phase 2 — Enrich (probe & fingerprint)

httpx — the workhorse

cat subs.txt | httpx -silent -title -tech-detect -status-code -ip -cdn -jarm \
  -favicon -screenshot -srd /tmp/shots -rl 150 -jsonl > probes.jsonl

Key flags:

Probes: -title, -server, -td, -tech-detect, -status-code, -cl, -content-length, -location, -method, -cname, -ip, -asn, -cdn (annotate CDN/WAF), -jarm, -favicon, -bh, -body-hash, -hh, -header-hash.
-screenshot -srd /path — Chromium-driven full-page captures written to <srd>/screenshot/<host>.png plus an <host>.txt with metadata. Trims disk fast.
-nf, -no-fallback — emit both HTTP and HTTPS, not just one. Pair with -ports http:80,https:443,http:8080,https:8443 to enumerate non-standard ports.
Rate: -t 50 (threads), -rl 150 (req/s, default 150), -rlm per-minute, -delay 200ms, -timeout 10, -maxhr max errors per host.
Match/filter: -mc 200,302, -mlc, -mwc, -fc, -fl, etc.
Burp / OpenAPI ingest: -l burp-export.xml -im burp, also -im jsonl|yaml|openapi|swagger.
TLS: -tls-probe, -tls-impersonate, -tls-grab, -min-tls-version.
-fr, -follow-redirects, -fhr, -follow-host-redirects.
-pd, -dashboard / -pdu, -dashboard-upload <jsonl> — upload results to PDCP. -auth configures the API key; the env ENABLE_CLOUD_UPLOAD=true turns this on by default.
-fep, -filter-error-page — emits a filtered_error_page.json side-file and removes soft-404s from the main output.
-sf, -secrets-file — runs a YAML rule set against responses for obvious secret patterns.

Output schema (-jsonl) — common keys: url, input, host, port, scheme, status_code, content_length, title, webserver, tech (array), cdn_name, cdn_type, ip, cname, a (array), aaaa, asn, jarm, favicon, body_hash, header_hash, time (response time), screenshot_path. Feed the same JSONL straight into nuclei v3.2 via -im jsonl.

Footguns:

-screenshot writes one file per host. 50k subdomains = 50k PNGs; this will fill /tmp. Use -srd to point at a real disk and prune.
-tech-detect is wappalyzer-based — accuracy is "best-effort", not authoritative. Always cross-check before reporting "site uses WordPress".
-cdn is required upstream of nuclei when scanning shared infrastructure (Cloudflare, CloudFront). Without it, nuclei templates hit the CDN's response, not the origin, and you get phantom findings.
Make sure you're on at least httpx v1.6.7 for -auth / -pd; older builds silently drop the upload flag.

dnsx — fast DNS toolkit

cat subs.txt | dnsx -silent -a -resp -cdn -asn -rl 1000 -auto-wildcard

Queries: -a (default), -aaaa, -cname, -ns, -txt, -srv, -ptr, -mx, -soa, -axfr, -caa, -any.

Probes: -cdn (annotate CDN name), -asn.

Rate/Resolvers: -t 100 threads, -rl -1 (disabled by default — always set this for big lists), -retry 2, -r resolvers.txt, -stream (no resume/stats/wildcard, for low-mem scans).

Bruteforce: dnsx -d target.com -w wordlist.txt (use placeholders like WORD.target.com for prefixed wordlists).

Wildcard handling:

-wd target.com -wt 5 — single-domain wildcard detection with threshold (default 5).
-auto-wildcard — new in 2024+: detects wildcard roots across mixed-domain input automatically. Mutually exclusive with -wd.
The PTR-on-CIDR / PTR-on-ASN trick: echo 173.0.0.0/24 | dnsx -ptr is a fast way to derive subdomains from a known range.

Footguns:

Default -r resolvers can rate-limit your IP fast on big runs. A trusted resolver list is half the battle.
Trace mode (-trace) recurses up to -trace-max-recursion 32767 by default — fine for one host, not for 10k.
-resume is incompatible with -stream mode.

tlsx — TLS / cert intel

cat subs.txt | tlsx -silent -san -cn -jarm -ja3 -json -mode auto -pre-handshake

Key flags:

-u / -l — input.
-san -cn — extract Subject Alternative Names and Common Name from the cert (best subdomain pivot you'll get out of the recon phase).
-jarm — JARM TLS fingerprint (passive infra clustering).
-ja3 — JA3 fingerprint.
-mode {auto|ctls|ztls|openssl} — TLS stack. ctls is Go's crypto/tls, ztls is zcrypto/tls (faster, capped at TLS 1.2), openssl shells out, auto (default) tries ctls then ztls.
-pre-handshake — terminate after serverhello; ~2x faster, but forces ztls mode (so TLS 1.2 ceiling).
-min-version tls12 -max-version tls13, -ci, -cipher-input — cipher / version sweeps for misconfig hunting.
-expired -self-signed -mismatched -revoked — flag misconfigured certs in output.

Footguns:

-pre-handshake won't see TLS 1.3 servers correctly (zcrypto cap).
-jarm is one of many fingerprints — cluster by JARM + favicon hash + server header for higher confidence.
The cert SANs are an under-used subdomain source — pipe tlsx -san back into dnsx.

Phase 3 — Detect (find vulns)

nuclei — template-driven vulnerability scanner

nuclei -l targets.jsonl -im jsonl -severity critical,high \
  -t cves/,exposures/ -c 25 -bs 25 -rl 150 -jsonl -o nuc.jsonl -auth

Current generation: nuclei v3.x. v2 is end-of-life on 2026-03-01 (nuclei-action@v2 and the legacy template directory layout). Migrate now if you haven't.

Key flags:

Targets: -u, -l, -iv 4|6, -sa, -scan-all-ips, -eh exclude hosts, -resume <file>.
Templates: -t path/, -w workflow.yaml, -tags cve,rce, -itags <tag> (include tags marked dangerous by default), -severity critical,high,medium,low,info, -tc <DSL> (template condition), -as, -automatic-scan, -disable-clustering.
Trust: -sa (signed-author), -cs (community signed), -dut, -disable-unsigned-templates. Use -sign to sign your own templates with NUCLEI_SIGNATURE_PRIVATE_KEY.
Rate: -c 25 (concurrent templates), -bs 25 (targets per template in bulk), -rl 150 (req/s), -rlm per-minute, -timeout 10, -retries 1, -ldp leave-default-ports, -mhe 30 max-host-error.
I/O: -jsonl, -o, -sresp / -srd (store request/response), -stats / -si 5 / -mp 9092 (metrics on localhost), -cup, -cloud-upload to PDCP.
Update: -update-templates (regularly!), -disable-update-check (-duc, deprecated in v3.7+ — replaced by -force-update-check and a cached state file at ~/.cache/nuclei/update-state.gob).
Auth: -auth (configures PDCP_API_KEY via prompt or -auth <key> in v3.2+), -tid team-id, -cup, -aname asset-name.
Interactsh: -iserver oast.live,oast.pro,… (CSV of servers), -itoken (self-hosted), -no-interactsh to disable OAST.
Fuzzing (v3.2+): -ft replace|prefix|postfix|infix, -fm multiple|single, -im jsonl|burp|openapi|swagger|yaml|postman to ingest HTTP traffic from httpx/proxify/Burp.
Uncover bridge: -uc -uq '<query>' -ue shodan,censys to fan out a template over an uncover result set.
Authenticated scans (v3.2+): -sf, -secret-file auth.yaml plus -ps prefetch-secrets. The same secrets file works across all existing templates without per-template edits.

Rate-limit tuning (from the PD optimization guide):

Hardware	`-c`	`-bs`	`-rl`	`-timeout`	`-retries`
1 GB / 1 CPU	2	2	10	30	3
4 GB / 4 CPU	10	10	50	20	2
8 GB / 4–8 CPU	25	25	200	15	1
32 GB+ / 16+	100	100	1000+	10	1

For bug-bounty programs that demand throttling (Intigriti, parts of Bugcrowd), drop to -rl 3 -c 2 and document the values in your report.

Footguns:

"9000 templates" doesn't mean "9000 distinct vulns" — many are variants, language localizations, or detection-only. Use -tags cve,rce + -severity critical,high first; expand if quiet.
Templates with weak matchers cause false positives in waves (issue #12956 was a single bad day in the public set). When a template lights up, verify the match yourself before reporting.
-stop-at-first-match is per-template, not global — and can break workflows that depend on multiple matches.
Default templates include some DoS-shaped ones (fuzz / dos tags) excluded from default scans; only run with -itags if you have permission.
-code is gated behind -enable-self-contained and a signed template by default; that's the safety net for the code protocol introduced in v3.0+.

interactsh — OOB interaction server

interactsh-client -v -server oast.live          # use hosted (default)
interactsh-client -server interact.mydomain.com -token MY-TOKEN -v   # self-hosted

The hosted servers rotate (oast.pro,oast.live,oast.site,oast.online, oast.fun,oast.me). For long-running scans or callbacks containing sensitive metadata, self-host.

Self-host recipe (DigitalOcean / AWS Lightsail / Hetzner — any $5/mo box works):

Buy a domain. In your registrar, add ns1 and ns2 glue records pointing to the VPS IP (Namecheap → Advanced DNS → "Personal DNS Server", or "Add nameserver"). Then change the domain's NS records to ns1.<your-domain>, ns2.<your-domain>.
Open ports on the VPS: 53/udp, 80, 443, 25, 587, 465, optionally 8443 for the web client. LDAP needs 389; SMTPS needs 465.
go install github.com/projectdiscovery/interactsh/cmd/interactsh-server@latest

Run via systemd:

ExecStart=/root/go/bin/interactsh-server \
  -domain interact.mydomain.com \
  -t SOME-RANDOM-TOKEN \
  -cidl 5 -cidn 6 \
  -hd /opt/interact-payloads/

-cidl/-cidn shorten the correlation-id (the random subdomain payload prefix) for cleaner OOB strings.

From nuclei: -iserver interact.mydomain.com -itoken SOME-RANDOM-TOKEN.

Footguns:

The hosted server's correlation IDs are long (c2*...yyyyyn.oast.pro). Some WAFs and SIEMs flag those as suspicious payloads. Self-hosting with -cidl 5 -cidn 6 produces shorter, less obvious strings.
Don't put your interactsh token in shell history — set it as an env var (INTERACTSH_TOKEN) and pass -itoken $INTERACTSH_TOKEN.

cvemap — CLI over CVE databases

cvemap -auth                       # one-time PDCP setup
cvemap -kev -severity critical -product nginx,apache -l 10
cvemap -cs '>8' -es '>0.5' -poc -re -j | jq -r '.[].id' | nuclei -id -

Note: the github repo is now projectdiscovery/cvemap → vulnx (rebranded). Both names work; vulnx adds richer filtering (--tags, --vuln-type, --vuln-age "<30", --term-facets, --range-facets, --detailed).

Key flags:

Filters: -id, -v/-vendor, -p/-product, -eproduct, -s/-severity, -cs/-cvss-score '>7', -es/-epss-score '>0.5', -ep/-epss-percentile, -age '<30', -a/-assignee, -vs/-vstatus {modified|confirmed|rejected|new|unconfirmed}.
Toggles (default true): -kev (CISA Known Exploited Vulnerabilities), -t/-template (has a nuclei template), -poc (has public POC), -h1/-hackerone (reported on H1), -re (remotely exploitable).
Output: -f field1,field2, -fe exclude_field, -lsi list-ids, -l 50, -offset, -j/-json, -q '<text>' free-text search, -epk page-keys for interactive paging.

Footguns:

Free-tier query budget per day on the PDCP CVE API — bulk pulls hit limits. Use -q instead of repeated single-CVE queries.
The -template true filter is the fastest way to narrow to "things I can scan today with nuclei".

Phase 4 — Notify & Plumbing

notify — multi-platform alerting

subfinder -d target.com -silent | notify -bulk -id recon
nuclei -l targets -severity critical -o nuc.txt; notify -data nuc.txt -bulk -provider slack,discord

Key flags:

-bulk — assemble all input into one message (chunked at -cl 4000 by default; some providers chunk lower).
-data <file> — input file (or pipe to stdin).
-p, -provider slack,discord,… — restrict providers.
-id recon,vulns — restrict to provider entries with these ids.
-mf, -msg-format 'Hey {{data}}' — format string with helpers {{datetime}}, {{date}}, {{time}}, {{count}}, {{data}}.
-d, -delay 2 — seconds between sends (avoid rate-limiting on Discord webhooks).
-rl, -rate-limit 1 — alternative rate cap.
-pc, -provider-config notify.yaml — alternate provider config.
-proxy http://… — route through your own proxy for auditability.

Provider config (~/.config/notify/provider-config.yaml):

slack:
  - id: recon
    slack_username: notify-bot
    slack_webhook_url: $SLACK_WEBHOOK   # env-var substitution supported
discord:
  - id: vulns
    discord_username: nuclei
    discord_webhook_url: $DISCORD_WEBHOOK
telegram:
  - id: vulns
    telegram_api_key: $TELEGRAM_BOT_TOKEN
    telegram_chat_id: "-1001234567890"
smtp:
  - id: digest
    smtp_server: smtp.gmail.com
    smtp_port: 587
    use_starttls: true
    smtp_password: $SMTP_PASSWORD
gotify:
  - id: ops
    gotify_host: https://gotify.internal
    gotify_token: $GOTIFY_TOKEN

Supported providers: Slack, Discord, Telegram, Pushover, Email (SMTP), Microsoft Teams, Google Chat, Gotify, Custom Webhook.

Footguns:

Webhook URLs are capability secrets — anyone with the URL can post to your channel. Don't put them in command lines (shell history, ps output, CI logs). Always use env-var substitution.
Discord rate-limits aggressively per webhook; -d 2 -cl 1800 is a safe default for bulk firehose runs.
Telegram bot tokens leak the bot's identity — rotate if you accidentally commit the config file. git-secrets or trufflehog --only-verified saves you here.
-bulk concatenates lines into a single message; for line-per- alert behavior (the default), omit -bulk.

Utilities

mapcidr — CIDR math

mapcidr -cidr 173.0.84.0/24 -sbc 10                  # split into 10-host chunks
cat ips.txt | mapcidr -a                              # aggregate IPs to CIDR
cat ips.txt | mapcidr -aa                             # approximate aggregation
mapcidr -cl 'ips.txt,cidrs.txt' -f4                    # combine lists, filter IPv4

Flags: -cidr/-cl, -sbc 10 (split by count), -sbh 1024 (split by host count), -a/-aggregate, -aa/-aggregate-approx, -f4/-f6 (filter), -count. Faster + more correct than ad-hoc Python for prepping naabu input lists.

cdncheck — CDN/WAF detection

cat ips.txt | cdncheck -resp -silent       # tag each IP with provider

Standalone CLI plus a Go library many tools (naabu, httpx, metabigor) embed. Knows Cloudflare, Akamai, Incapsula, Sucuri, Fastly, CloudFront, plus less common players. Use it when you need a reason to exclude IPs from a port scan, beyond naabu's built-in -exclude-cdn.

proxify — Swiss army proxy

proxify -o logs/ -dump-req -dump-resp                  # HTTP :8888 / SOCKS5 :10080
proxify -dm 'api.target.com:127.0.0.1'                  # DNS mapping
proxify -mf 'host == "api.target.com"' -req-mf 'remove_header("X-Auth")'

Key flags:

-http-addr, -socks5-addr — listen addresses.
-o logs/ — JSONL-per-request log dir; -dump-req, -dump-resp.
-dm domain:ip,domain:ip — DNS mapping (override resolution).
-mf <DSL> / -req-mf / -resp-mf — match filter + match-and- replace via DSL.
-cert, -key, -out-cert-path — custom MITM CA for transparent TLS interception.
-elastic-url, -kafka-broker — stream into your SIEM.
-upstream-proxy — chain in front of Burp/Caido.

Use cases: full-fidelity request log for compliance / audit; sit between Burp and the target to add structured logging; off-the-shelf DNS rebinding via -dm; replay traffic into nuclei v3.2 (nuclei -im jsonl -input proxify.jsonl).

simplehttpserver — Go take on Python's `http.server`

simplehttpserver -listen 0.0.0.0:8000 -upload -basic-auth user:pass -path /opt/share
simplehttpserver -tls -domain target.example.com -cert cert.pem -key cert.key
simplehttpserver -tcp 0.0.0.0:4444 -tls -rules tls-banner.yaml      # full TCP server

Key flags: -listen, -path, -https / -cert / -key, -basic-auth user:pass -realm "…", -upload -max-file-size 100, -cors, -sandbox, -tcp, -rules <yaml>, -verbose.

Use it for: staging RCE payloads (curl --upload-file shell.bin), serving SSRF callback content, running a quick TLS-banner emulator when you're testing detection rules.

aix — LLM CLI

aix is a thin CLI wrapper to call LLM APIs from your terminal — useful for one-shot prompts during recon ("classify these 200 titles into categories"), but it's intentionally simple. If you want template-driven AI workflows tied to scan results, you're really looking at PDCP's template editor or nuclei v3.2's AI template generation, not aix.

Pipeline Cookbook

The recipes from the original SKILL.md, plus modern additions.

1. Classic external recon → vuln scan

subfinder -d target.com -silent \
  | dnsx -silent -resp -rl 1000 \
  | httpx -silent -title -tech-detect -status-code -jsonl \
  | nuclei -im jsonl -severity critical,high -t cves/

2. Cloud asset → vuln scan

cloudlist -silent -exclude-private -json \
  | jq -r '.host' \
  | httpx -silent \
  | nuclei -t cves/

3. ASN / network fingerprinting

subfinder -d target.com -silent | httpx -silent -asn -jsonl | jq -c '{host, asn}'

4. JARM TLS clustering (find sibling infra)

subfinder -d target.com -silent | httpx -silent -jarm -jsonl \
  | jq -r 'select(.jarm) | "\(.url)\t\(.jarm)"' | sort -k2 | uniq -f1 -c

5. Favicon hash hunting

subfinder -d target.com -silent | httpx -silent -favicon
# then pivot the hash:
uncover -q 'http.favicon.hash:-1234567890' -e shodan

6. Visual recon

subfinder -d target.com -silent | httpx -silent -screenshot -srd /tmp/shots
# review with: feh /tmp/shots/screenshot/ -F

7. Port-scan-driven probing

sudo naabu -host target.com -top-ports 1000 -exclude-cdn -silent \
  | httpx -silent -ports - \
  | nuclei -t exposures/

8. Real-time alerting layer

nuclei -l targets.txt -severity critical,high -silent \
  | notify -bulk -id vulns

9. SPA crawl (modern, JS-heavy targets)

katana -u https://app.target.com -headless -jc -kf all -d 5 -c 10 -p 3 -jsonl \
  | jq -r '.endpoint' \
  | httpx -silent -mc 200,401,403 \
  | nuclei -t exposures/,vulnerabilities/

10. KEV-filtered nuclei (only run templates for currently-exploited CVEs)

cvemap -kev -template -j -silent | jq -r '.[].id' | sort -u > kev-cves.txt
nuclei -l targets.txt -id "$(paste -sd, kev-cves.txt)" -severity critical,high

11. ASN sweep via uncover

uncover -q 'asn:"AS13335"' -e shodan,censys -uf 'host:port' -ul 500 -silent \
  | httpx -silent -title -tech-detect

12. Bug-bounty-friendly throttled chain (program demands rate limiting)

subfinder -d target.com -rl 5 -silent \
  | dnsx -silent -rl 50 \
  | httpx -silent -rl 5 -rlm 60 \
  | nuclei -rl 3 -c 2 -severity critical,high

13. Cert-pivot horizontal recon

tlsx -u target.com -san -cn -resp-only -silent \
  | sort -u | dnsx -silent -resp \
  | httpx -silent

14. Authenticated nuclei scan (v3.2+)

# auth.yaml:
# static:
#   - type: cookie
#     domains: ["target.com"]
#     cookies:
#       - key: session
#         value: $SESSION_COOKIE
nuclei -l targets.txt -sf auth.yaml -ps -severity high,critical

15. ProjectDiscovery Cloud handoff

httpx -auth                              # one-time
httpx -l subs.txt -silent -pd -aname "target.com-prod"   # upload as a named asset group
nuclei -list targets.txt -cloud-upload   # results appear in the dashboard

Common Footguns (cross-cutting)

What happens	Why it bites	How to avoid
`subfinder -all` returns 0 / very few	No provider keys configured; the free sources alone are sparse	Configure ≥5 free-tier providers; verify with `subfinder -ls`
`httpx -screenshot` fills `/tmp`	One PNG per host, no automatic rotation	`-srd /large/disk`; prune per-host older than N days
naabu against `*.cloudflare-protected` is useless	You're scanning the CDN's IP pool, not the origin	`-exclude-cdn` (and validate with `cdncheck` upstream)
nuclei against shared CDN hosts thrashes neighbors	Rate-limit hits multiple tenants in the same /24	`-rl 3 -c 2` for bug-bounty; pair with `httpx -cdn`
FOFA / Censys "no API key defined"	Wrong config format — they need `email:key` / `id:secret`	Single string is the key; use the `:` separator
`dnsx` saturates resolvers	Default rate is unlimited; default `-r` resolvers throttle hard	Always set `-rl`; bring your own resolvers list
`chaos -d target.com` returns nothing	Missing `PDCP_API_KEY` — chaos requires auth, no free unauth path	Get key from `cloud.projectdiscovery.io`; export env var
Webhook URL in shell history	Anyone reading `~/.bash_history` (or your tmux scrollback) can post	Use env-var substitution in `provider-config.yaml`
katana headless OOMs the box	Hundreds of Chromium tabs spawned in parallel	Cap `-p 5 -c 10` first; scale only after watching `top`
Old `nuclei-templates` produce FPs	Templates updated daily; a stale clone misses fixes	Run `nuclei -update-templates` weekly
"interact.sh callback in client logs"	Hosted interactsh sees your payload metadata	Self-host for sensitive engagements
Self-XSS template fires on benign reflection	Generic word matchers, no protocol gating	Pre-condition templates with method + path + tech signals
Pipeline drops blank lines silently	`subfinder	httpx` swallows lines with no protocol prefix

Provider Auth & Secrets

Tool/source	Config path	Env var(s)	Notes
All PD tools	n/a	`PDCP_API_KEY`	Cloud upload, chaos, cvemap, nuclei `-auth`
subfinder	`~/.config/subfinder/provider-config.yaml`	`SUBFINDER_*` per source	`subfinder -ls` lists all sources
chaos	n/a	`PDCP_API_KEY`	Also accepts `-key XXX`
uncover	`~/.config/uncover/provider-config.yaml`	`SHODAN_API_KEY`, `CENSYS_API_ID`+`CENSYS_API_SECRET`, `FOFA_EMAIL`+`FOFA_KEY`, `QUAKE_TOKEN`, `HUNTER_API_KEY`, `ZOOMEYE_API_KEY`, `NETLAS_API_KEY`, `CRIMINALIP_API_KEY`	Env vars take precedence over YAML
cloudlist	`~/.config/cloudlist/provider-config.yaml`	`AWS_`, `GOOGLE_APPLICATION_CREDENTIALS`, `AZURE_`	AWS supports cross-account `assume_role_arn` + `external_id`
notify	`~/.config/notify/provider-config.yaml`	`SLACK_WEBHOOK`, `DISCORD_WEBHOOK`, `TELEGRAM_BOT_TOKEN`, `SMTP_PASSWORD`, `GOTIFY_TOKEN`	Use `$ENV` substitution inside the YAML
interactsh-server	`~/.config/interactsh-server/config.yaml`	n/a	`-t TOKEN` passed at startup; client passes `-token TOKEN`
nuclei v3.2 auth	per-template `auth.yaml` (passed via `-sf`)	per-template variables	Future: HashiCorp Vault / AWS Secrets Manager / 1Password

Secret discipline:

chmod 0600 ~/.config/*/provider-config.yaml.
Add .config/ and *-config.yaml to ~/.gitignore_global.
Never cat or printenv a config in screenshots or asciinema.
Run trufflehog --only-verified filesystem ~/.config after a major refactor — it catches accidental key duplicates.
Rotate the webhook tokens after any leaks; rotate the PD API key if you publish from a CI runner that other people can read.

Modern Integration Patterns

Nuclei v3 → v3.2 migration cheatsheet

-json → -jsonl (the old -json was a JSON array, not lines).
-disable-update-check / -duc is deprecated in v3.7+; new flag is -force-update-check, and an update-state cache lives at ~/.cache/nuclei/update-state.gob.
New protocols: code: (gated behind -enable-self-contained + signed templates), javascript: (LDAP brute-force, Kerberos, custom checks), multi-protocol.
New template features: flow: (orchestrate multiple protocols inside one template), pre-condition: (skip the template unless DSL matches first), self-contained: true (no host required — use for cloud-config-review templates).
Fuzzing: -ft replace|prefix|postfix|infix|replace-regex, -fm multiple|single, plus -im ingest from Burp/JSONL/OpenAPI/ Swagger/Postman.
Authenticated scans: -sf auth.yaml -ps. Works with every existing template — no template edits required.
Signing: -sign with NUCLEI_SIGNATURE_PRIVATE_KEY; unsigned templates can be blocked with -dut, -disable-unsigned-templates.

ProjectDiscovery Cloud Platform (PDCP)

One API key (PDCP_API_KEY) authenticates all PD tools.
<tool> -auth is a one-time interactive setup; persists the key.
httpx -pd/-dashboard uploads probe results to the cloud dashboard. -aname groups them as named assets.
nuclei -cloud-upload (or -cup) uploads scan results. -tid TEAM-ID targets a specific team.
cvemap requires -auth to query the CVE API at all.
Always set ENABLE_CLOUD_UPLOAD=true only if you genuinely want every run to upload — engagement data leaks otherwise.

PD CLI vs BBOT

Use PD CLI when…	Use BBOT when…
You're hunting actively — composing pipelines in a shell	You want hands-off attack-surface monitoring
You want grep-able JSONL you can diff between runs	You want a unified asset graph + module DAG
You're integrating with proxies / Burp / Caido	You want one tool that does everything end-to-end
You need fine rate-limit control per stage	You want sensible defaults you don't think about
You're writing custom Nuclei templates	You want OSINT + cloud + DNS + web in one binary

They're complementary, not exclusive — most operators run BBOT for continuous discovery and the PD pipeline for targeted hunting.

Burp / Caido handoffs

katana -hl -jsonl → import into Burp via proxify as upstream; Burp's history is now the crawl plus the manual replays.
proxify -mf 'contains(path, "/admin")' filters interesting traffic out of a noisy Burp session; pipe the JSONL into nuclei v3.2 via -im jsonl.
Burp's project file → nuclei -im burp -input file.xml re-runs Burp-captured requests as fuzzing seeds.

JSONL → SIEM / report

All PD tools output a common-ish JSONL schema with url/host/ port/status_code/title/tech/time.
notify -data results.jsonl -bulk -mf '{{datetime}} {{data}}' is the lazy way; proxify -elastic-url … or -kafka-broker … is the production way.
For Splunk/Elastic: use httpx/nuclei -store-resp -srd resp/ to capture full bodies, then ship resp/ as evidence alongside the JSONL index.

Sources

Upstream PD documentation and blog posts that informed each section. URLs grouped by tool so a reader can verify any specific claim.

pdtm: https://github.com/projectdiscovery/pdtm • https://docs.projectdiscovery.io/opensource/pdtm/usage • https://docs.projectdiscovery.io/quickstart • https://projectdiscovery.io/blog/getting-started-with-projectdiscovery-in-linux-and-windows

subfinder: https://github.com/projectdiscovery/subfinder • https://docs.projectdiscovery.io/opensource/subfinder/install • https://docs.projectdiscovery.io/opensource/subfinder/running • https://projectdiscovery.io/blog/do-you-really-know-subfinder-an-in-depth-guide-to-all-features-of-subfinder-beginner-to-advanced • https://highon.coffee/blog/subfinder-cheat-sheet • https://github.com/projectdiscovery/subfinder/discussions/1126 • https://github.com/projectdiscovery/subfinder/discussions/1563

naabu: https://github.com/projectdiscovery/naabu • https://docs.projectdiscovery.io/opensource/naabu/usage • https://docs.projectdiscovery.io/opensource/naabu/running • https://docs.projectdiscovery.io/opensource/naabu/overview

katana: https://github.com/projectdiscovery/katana • https://projectdiscovery.io/blog/introducing-katana-the-best-cli-web-crawler • https://projectdiscovery.io/blog/a-deep-dive-on-katana-field-extraction

cloudlist: https://docs.projectdiscovery.io/opensource/cloudlist/overview • https://docs.projectdiscovery.io/opensource/cloudlist/providers • https://docs.projectdiscovery.io/opensource/cloudlist/running • https://docs.projectdiscovery.io/opensource/cloudlist/usage • https://docs.projectdiscovery.io/cloud/integrations

chaos: https://chaos.projectdiscovery.io • https://chaos.projectdiscovery.io/docs/api-key • https://chaos.projectdiscovery.io/docs/fetch-subdomains • https://github.com/projectdiscovery/chaos-client

uncover: https://docs.projectdiscovery.io/opensource/uncover/usage • https://projectdiscovery.io/blog/uncover • https://github.com/projectdiscovery/nuclei-docs/blob/main/docs/nuclei/get-started.md

alterx: https://docs.projectdiscovery.io/opensource/alterx/usage • https://docs.projectdiscovery.io/opensource/alterx/running • https://projectdiscovery.io/blog/introducing-alterx-simplifying-active-subdomain-enumeration-with-patterns

httpx: https://github.com/projectdiscovery/httpx • https://docs.projectdiscovery.io/opensource/httpx/overview • https://docs.projectdiscovery.io/opensource/httpx/running • https://docs.projectdiscovery.io/opensource/httpx/usage • https://projectdiscovery.io/blog/introducing-httpx-dashboard-2 • https://highon.coffee/blog/httpx-cheat-sheet

dnsx: https://github.com/projectdiscovery/dnsx • https://docs.projectdiscovery.io/opensource/dnsx/overview • https://docs.projectdiscovery.io/opensource/dnsx/usage

tlsx: https://projectdiscovery.io/blog/a-hackers-guide-to-ssl-certificates-featuring-tlsx

nuclei: https://github.com/projectdiscovery/nuclei • https://docs.projectdiscovery.io/opensource/nuclei/running • https://projectdiscovery.io/blog/ultimate-nuclei-guide • https://projectdiscovery.io/blog/nuclei-3-2 • https://projectdiscovery.io/blog/scanning-login-protected-targets-with-nuclei • https://projectdiscovery.io/blog/nuclei-fuzzing-for-unknown-vulnerabilities • https://blog.projektckmt.com/blog/optimizing-nuclei-mass-scanning-strategies-configurations • https://github.com/projectdiscovery/nuclei-action/discussions/104 • https://github.com/projectdiscovery/nuclei/issues/6672

interactsh: https://github.com/projectdiscovery/interactsh • https://docs.projectdiscovery.io/opensource/interactsh/server • https://docs.projectdiscovery.io/opensource/interactsh/running • https://projectdiscovery.io/blog/nuclei-interactsh-integration • https://projectdiscovery.io/blog/interactsh-v1

cvemap: https://github.com/projectdiscovery/cvemap • https://docs.projectdiscovery.io/opensource/cvemap/usage • https://docs.projectdiscovery.io/opensource/cvemap/running • https://projectdiscovery.io/blog/announcing-cvemap-from-projectdiscovery • https://projectdiscovery.io/blog/future-of-automating-nuclei-templates-with-ai

notify: https://github.com/projectdiscovery/notify • https://docs.projectdiscovery.io/opensource/notify/overview • https://docs.projectdiscovery.io/opensource/notify/running • https://docs.projectdiscovery.io/opensource/notify/provider-config • https://github.com/projectdiscovery/notify/releases

mapcidr / cdncheck / proxify / simplehttpserver / aix: https://docs.projectdiscovery.io/opensource • https://projectdiscovery.io/blog/projectdiscovery-best-kept-secrets • https://projectdiscovery.io/blog/proxify-portable-cli-based-proxy • https://github.com/projectdiscovery/simplehttpserver • https://github.com/projectdiscovery/aix

Cross-cutting / PDCP / templates: https://docs.projectdiscovery.io/opensource • https://docs.projectdiscovery.io/templates/faq • https://projectdiscovery.io/blog/aws-cloud-security-config-review-using-nuclei-templates • https://projectdiscovery.io/blog/gcp-cloud-configuration-review-templates-nuclei-templates-v10-2-0 • https://projectdiscovery.io/blog/building-one-shot-recon

name	project-discovery-workflow
description	ProjectDiscovery toolchain operator reference. 20 CLI tools (pdtm tool manager, subfinder, naabu, katana, cloudlist, chaos, uncover, alterx, httpx, dnsx, tlsx, nuclei, interactsh, cvemap, notify, mapcidr, cdncheck, proxify, simplehttpserver, aix) organized as a 4-phase pipeline (Discover, Enrich, Detect, Notify) chained via stdin/stdout. Covers current flags, footguns, provider-config.yaml auth (PDCP_API_KEY, SHODAN_API_KEY, CENSYS_API_ID/SECRET, FOFA_EMAIL/KEY, QUAKE_TOKEN, HUNTER_API_KEY, ZOOMEYE_API_KEY), Nuclei v3 changes (-jsonl, -duc deprecation, -auth, flow/code protocols, self-contained templates, authenticated scanning via -secret-file), rate-limit tuning matrices, modern chains (SPA crawl with katana -headless, KEV-filtered nuclei via cvemap, JARM cert clustering with tlsx, ASN sweep via uncover, PDCP cloud upload with -pd / -auth), interactsh self-hosting on a VPS (NS records + -domain + -t token), notify provider-config webhooks, and PD-vs-BBOT positioning. Use when wiring up bug-bounty or pentest recon pipelines, troubleshooting subfinder providers, tuning nuclei -c/-bs/-rl/-rlm for the target box, or piping output into Slack/Discord/Jira via notify.
sources	docs.projectdiscovery.io, github.com/projectdiscovery, projectdiscovery.io/blog
report_count	0

ProjectDiscovery Toolchain — Operator Reference

When to reach for PD vs a framework like BBOT: PD wins when you want a transparent pipeline you read top-to-bottom in your shell history — small Unix tools you can swap in and out, with output you can grep and diff. BBOT wins when you want one tool that orchestrates a hundred modules and produces a unified asset graph. Most operators end up running both — BBOT for hands-off attack-surface monitoring, PD tools when actively hunting.

pdtm — Tool Manager

Use pdtm for everything. Skipping it and go install-ing each tool individually is the most common source of "I have nuclei v2, why don't templates resolve?" pain.

go install -v github.com/projectdiscovery/pdtm/cmd/pdtm@latest  # needs Go 1.24.3+
pdtm -ia          # install all PD tools
pdtm -ua          # update all
pdtm -i nuclei,httpx,katana   # install a subset
pdtm -sp          # show the binary path
pdtm -up          # self-update pdtm itself

Phase 1 — Discover (map attack surface)

subfinder — passive subdomain enumeration

subfinder -d target.com -all -recursive -silent | tee subs.txt

Key flags (full list via -h):

-d / -dL — single domain or file of domains.
-all — uses every configured source (slow, more dupes; default uses a curated subset).
-recursive — only sources that can recurse on found subdomains.
-active — DNS-resolve discovered names; required for -oI (include IP) and -nW (filter wildcards).
-pc, -provider-config — point at a different provider-config.yaml.
-s, -sources / -es, -exclude-sources — narrow or block sources; subfinder -ls lists the ~30 supported.
-rl — HTTP requests per second across sources.
-t — concurrent resolution goroutines (only used with -active).
-oJ, -json — JSONL output. Combine with -cs to include source attribution per record.
-r / -rL — custom resolvers (file or comma list).

Footguns:

-all without API keys returns a lot of junk from the noisier free sources. Configure providers first, then -all is actually useful.
Default resolvers (8.8.8.8 etc.) get rate-limited fast on big sweeps. Use your own resolver list (trickest/resolvers, proabiral/Fresh-Resolvers) and update it weekly.
DNS caching at the source side can hide newly added subdomains — passive enum is a leading indicator, not a real-time signal.
subfinder -d example.com -oI -active returns nothing if -active isn't paired with -nW; the IP-include path only fires on resolved records (see GitHub discussion #1126).

Provider auth (~/.config/subfinder/provider-config.yaml):

Free-tier keys worth getting: chaos, virustotal, securitytrails, censys (250/mo), github, fullhunt, intelx, leakix, netlas.
FOFA wants email:key, censys wants api_id:api_secret, github wants raw PATs.
passivetotal uses email:password.
Multiple keys per source rotate automatically.

naabu — fast port scanner

sudo naabu -host target.com -top-ports 1000 -exclude-cdn -nmap-cli 'nmap -sV -oX scan.xml'

Key flags:

-host / -list / -eh / -ef — input, exclusions.
-p, -tp top-ports {100|1000|full}, -ep, -pf — port selection.
-s SYN|CONNECT — scan type. SYN needs root; otherwise it silently falls back to CONNECT.
-c 25 — worker threads; -rate 1000 — packets/sec.
-sV / -sV-fast / -sV-workers / -sV-probes — built-in service version detection via the nmap-service-probes file.
-ec, -exclude-cdn — for Cloudflare / Akamai / Incapsula / Sucuri IPs, only scan 80/443.
-cdn — annotate detected CDN in output.
-passive — use Shodan InternetDB (no packets sent).
-nmap-cli 'nmap -sV …' — pipe found ports into a real nmap.
-metrics-port — exposes JSON scan stats on localhost.
-iv 4,6 -sa — scan all v4+v6 IPs a host resolves to.

Footguns:

Default config assumes a VPS — local boxes need lower rates or you saturate your uplink and the scan looks "slow" while you're actually ddosing yourself.
-exclude-cdn is required for Cloudflare/Akamai etc. unless you want to fingerprint the CDN, not the origin. Without it, you scan the CDN's IP pool and waste a lot of packets for cloudflare:443 banner outputs.
Without sudo you get CONNECT scans, which are slower and noisier (full TCP handshake, more IDS hits).
Masscan goes faster but is less reliable; naabu is the right default for bug-bounty-scale lists where you re-probe with httpx anyway.

katana — next-gen crawler

katana -u https://target.com -d 5 -jc -ps -iqp -kf all -c 30 -ef css,woff,svg

Key flags:

-u / -list — input.
-d 5 — crawl depth.
-c 30 -p 10 -rl 100 -rlm 500 — concurrency, parallelism, rate limits (req/s and req/min).
-jc, -js-crawl — parse endpoints out of JS bundles (the killer feature on SPAs).
-hl, -headless — Chromium-backed crawl for React/Vue/Angular. Combine with -sc, -system-chrome to reuse the host Chrome, -ho, -headless-options for flags, -nos, -no-sandbox in containers, -page-load-strategy {normal|eager|none} to control the wait model.
-cs / -cos / -fs {dn|rdn|fqdn} / -ns / -do — scope control.
-ef css,txt,md extension filter; -ndef disables the default extension filter. -mr / -fr regex match/filter.
-kf all|robotstxt|sitemapxml|… — known-file harvesters.
-iqp, -ignore-query-params — collapses noisy parametric URLs.
-ps, -passive — pull URLs from waybackmachine/commoncrawl/alienvault.
-aff, -automatic-form-fill — fills forms during the crawl (experimental, worth turning on for endpoint discovery).
-jsonl — structured output you can feed back into nuclei/httpx.

Footguns:

Headless mode shells out to Chrome; on big lists you'll spawn hundreds of tabs and OOM. Cap with -p 5 -c 10 first, then scale.
DSL filters apply to jsonl keys, not the URL — -fr against the text URL works, -fr against a key needs the DSL pattern.
katana -u target.com -kf all against SPA-heavy targets without -headless will look "shallow" because most endpoints only emit after JS execution.

cloudlist — multi-cloud asset enumeration

cloudlist -provider aws,gcp,do -exclude-private -json | jq .

Key flags:

-pc, -provider-config — ~/.config/cloudlist/provider-config.yaml.
-p / -id / -s — filter by provider, by provider-config id, by service (e.g. -s ec2,route53,lambda).
-host / -ip / -ep, -exclude-private — output shape and pruning.

Footguns:

Cloudlist is the blue-team tool — it inventories your accounts, not someone else's. Don't put it on a recon checklist for an external target.
AWS supports assume_role_arn + external_id for cross-account scans; if the role isn't trusted to be assumed from your scanning account, the call silently returns zero assets.
GCP has two modes: per-service APIs (fast, project-scoped) and the Cloud Asset Inventory API (org-wide; needs cloudasset.assets.* perms).
Azure needs the explicit minimum role (see PD docs — VMs, NICs, Public IPs, DNS, Storage, AKS, CDN reads).
18 providers supported (aws, gcp, azure, alibaba, digitalocean, scaleway, heroku, linode, fastly, namecheap, cloudflare, hetzner, kubernetes, nomad, consul, terraform, openstack, ibm).

chaos — PD's internet-wide DNS dataset

chaos -d uber.com -silent | httpx -silent -title -tech-detect

Auth: PDCP_API_KEY env var, OR chaos -d uber.com -key XXX. Get the key from cloud.projectdiscovery.io. The old CHAOS_KEY env is gone — use PDCP_API_KEY.

REST: GET https://dns.projectdiscovery.io/dns/{domain}/subdomains with Authorization: API_KEY. Returns {"domain":"…","subdomains":["…"],"count":N}.

Footguns:

Chaos is a snapshot, not a live resolver. Subdomains that resolved hours ago may be stale; subdomains added in the last hour may be missing. Pair with subfinder and your own crt.sh poll.
Bug-bounty programs that publish in-scope hosts to Chaos are the highest-signal targets — chaos -d <program> first, then enrich.

uncover — query Shodan/Censys/FOFA/Quake/Hunter/ZoomEye/Netlas/CriminalIP/PublicWWW/HunterHow

uncover -q 'ssl:"target.com"' -e shodan,censys -ul 500 -uf 'host:port'

Key flags:

-q '…' — query (stdin/file/CLI).
-e shodan,censys,fofa,quake,hunter,zoomeye,netlas,criminalip,publicwww,hunterhow,shodan-idb — engine(s). Default is shodan.
-asq jira — use ProjectDiscovery's curated "awesome search queries".
-uf ip:port|host:port|ip|host — output field shape.
-ul 100 — limit (default 100).
-ur 60 — req/min when an engine doesn't expose its rate limit.
Engine-specific aliases (-s/-shodan, -sd/-shodan-idb, -ff/-fofa, -cs/-censys, -qk/-quake, -ht/-hunter, -ze/-zoomeye, -ne/-netlas, -cl/-criminalip).
Within nuclei: -uc -uq '…' -ue shodan to fan-out templates over uncover results.

Auth — environment variables (or ~/.config/uncover/provider-config.yaml):

SHODAN_API_KEY, CENSYS_API_ID+CENSYS_API_SECRET, FOFA_EMAIL+FOFA_KEY, QUAKE_TOKEN, HUNTER_API_KEY, ZOOMEYE_API_KEY, NETLAS_API_KEY, CRIMINALIP_API_KEY, PUBLICWWW_API_KEY, HUNTERHOW_API_KEY.

Footguns:

Each engine bills queries differently — Shodan free tier is 100 results/mo, Censys is 250/mo. -ul 500 against Shodan eats your monthly budget on a single command.
FOFA requires email:key, not just key; subfinder/uncover both fail silently with "no API key defined" when only the key is set (see subfinder issue #1563).
Shodan-idb is the free InternetDB — no key required, far less data than the paid Shodan engine.

alterx — DSL-based subdomain permutation

chaos -d tesla.com | alterx -enrich -p '{{sub}}-{{word}}.{{suffix}}' | dnsx -silent

Key flags:

-l / -p / -pp — input subdomains, pattern, payload (key=value).
-en, -enrich — auto-extract words from the input as payloads.
-es, -estimate — preview the permutation count before generating.
-limit / -ms 100mb / -o — bound output size.
-ac — point at a custom permutation config.

DSL variables (nuclei-style):

{{sub}} — the input subdomain string.
{{word}} — the payload word.
{{suffix}} — the TLD/eTLD.
{{root}} — the registered domain.

Footguns:

Alterx generates candidates, not results — pipe to dnsx -silent to validate. Without -enrich, you'll generate millions of permutations that mostly resolve to nothing.
The default permutation file lives at ~/.config/alterx/permutation_v0.0.1.yaml; if you've edited it and pdtm later overwrites it, alterx silently picks up the new file.

Phase 2 — Enrich (probe & fingerprint)

httpx — the workhorse

cat subs.txt | httpx -silent -title -tech-detect -status-code -ip -cdn -jarm \
  -favicon -screenshot -srd /tmp/shots -rl 150 -jsonl > probes.jsonl

Key flags:

Probes: -title, -server, -td, -tech-detect, -status-code, -cl, -content-length, -location, -method, -cname, -ip, -asn, -cdn (annotate CDN/WAF), -jarm, -favicon, -bh, -body-hash, -hh, -header-hash.
-screenshot -srd /path — Chromium-driven full-page captures written to <srd>/screenshot/<host>.png plus an <host>.txt with metadata. Trims disk fast.
-nf, -no-fallback — emit both HTTP and HTTPS, not just one. Pair with -ports http:80,https:443,http:8080,https:8443 to enumerate non-standard ports.
Rate: -t 50 (threads), -rl 150 (req/s, default 150), -rlm per-minute, -delay 200ms, -timeout 10, -maxhr max errors per host.
Match/filter: -mc 200,302, -mlc, -mwc, -fc, -fl, etc.
Burp / OpenAPI ingest: -l burp-export.xml -im burp, also -im jsonl|yaml|openapi|swagger.
TLS: -tls-probe, -tls-impersonate, -tls-grab, -min-tls-version.
-fr, -follow-redirects, -fhr, -follow-host-redirects.
-pd, -dashboard / -pdu, -dashboard-upload <jsonl> — upload results to PDCP. -auth configures the API key; the env ENABLE_CLOUD_UPLOAD=true turns this on by default.
-fep, -filter-error-page — emits a filtered_error_page.json side-file and removes soft-404s from the main output.
-sf, -secrets-file — runs a YAML rule set against responses for obvious secret patterns.

Footguns:

-screenshot writes one file per host. 50k subdomains = 50k PNGs; this will fill /tmp. Use -srd to point at a real disk and prune.
-tech-detect is wappalyzer-based — accuracy is "best-effort", not authoritative. Always cross-check before reporting "site uses WordPress".
-cdn is required upstream of nuclei when scanning shared infrastructure (Cloudflare, CloudFront). Without it, nuclei templates hit the CDN's response, not the origin, and you get phantom findings.
Make sure you're on at least httpx v1.6.7 for -auth / -pd; older builds silently drop the upload flag.

dnsx — fast DNS toolkit

cat subs.txt | dnsx -silent -a -resp -cdn -asn -rl 1000 -auto-wildcard

Queries: -a (default), -aaaa, -cname, -ns, -txt, -srv, -ptr, -mx, -soa, -axfr, -caa, -any.

Probes: -cdn (annotate CDN name), -asn.

Rate/Resolvers: -t 100 threads, -rl -1 (disabled by default — always set this for big lists), -retry 2, -r resolvers.txt, -stream (no resume/stats/wildcard, for low-mem scans).

Bruteforce: dnsx -d target.com -w wordlist.txt (use placeholders like WORD.target.com for prefixed wordlists).

Wildcard handling:

-wd target.com -wt 5 — single-domain wildcard detection with threshold (default 5).
-auto-wildcard — new in 2024+: detects wildcard roots across mixed-domain input automatically. Mutually exclusive with -wd.
The PTR-on-CIDR / PTR-on-ASN trick: echo 173.0.0.0/24 | dnsx -ptr is a fast way to derive subdomains from a known range.

Footguns:

Default -r resolvers can rate-limit your IP fast on big runs. A trusted resolver list is half the battle.
Trace mode (-trace) recurses up to -trace-max-recursion 32767 by default — fine for one host, not for 10k.
-resume is incompatible with -stream mode.

tlsx — TLS / cert intel

cat subs.txt | tlsx -silent -san -cn -jarm -ja3 -json -mode auto -pre-handshake

Key flags:

-u / -l — input.
-san -cn — extract Subject Alternative Names and Common Name from the cert (best subdomain pivot you'll get out of the recon phase).
-jarm — JARM TLS fingerprint (passive infra clustering).
-ja3 — JA3 fingerprint.
-mode {auto|ctls|ztls|openssl} — TLS stack. ctls is Go's crypto/tls, ztls is zcrypto/tls (faster, capped at TLS 1.2), openssl shells out, auto (default) tries ctls then ztls.
-pre-handshake — terminate after serverhello; ~2x faster, but forces ztls mode (so TLS 1.2 ceiling).
-min-version tls12 -max-version tls13, -ci, -cipher-input — cipher / version sweeps for misconfig hunting.
-expired -self-signed -mismatched -revoked — flag misconfigured certs in output.

Footguns:

-pre-handshake won't see TLS 1.3 servers correctly (zcrypto cap).
-jarm is one of many fingerprints — cluster by JARM + favicon hash + server header for higher confidence.
The cert SANs are an under-used subdomain source — pipe tlsx -san back into dnsx.

Phase 3 — Detect (find vulns)

nuclei — template-driven vulnerability scanner

nuclei -l targets.jsonl -im jsonl -severity critical,high \
  -t cves/,exposures/ -c 25 -bs 25 -rl 150 -jsonl -o nuc.jsonl -auth

Current generation: nuclei v3.x. v2 is end-of-life on 2026-03-01 (nuclei-action@v2 and the legacy template directory layout). Migrate now if you haven't.

Key flags:

Targets: -u, -l, -iv 4|6, -sa, -scan-all-ips, -eh exclude hosts, -resume <file>.
Templates: -t path/, -w workflow.yaml, -tags cve,rce, -itags <tag> (include tags marked dangerous by default), -severity critical,high,medium,low,info, -tc <DSL> (template condition), -as, -automatic-scan, -disable-clustering.
Trust: -sa (signed-author), -cs (community signed), -dut, -disable-unsigned-templates. Use -sign to sign your own templates with NUCLEI_SIGNATURE_PRIVATE_KEY.
Rate: -c 25 (concurrent templates), -bs 25 (targets per template in bulk), -rl 150 (req/s), -rlm per-minute, -timeout 10, -retries 1, -ldp leave-default-ports, -mhe 30 max-host-error.
I/O: -jsonl, -o, -sresp / -srd (store request/response), -stats / -si 5 / -mp 9092 (metrics on localhost), -cup, -cloud-upload to PDCP.
Update: -update-templates (regularly!), -disable-update-check (-duc, deprecated in v3.7+ — replaced by -force-update-check and a cached state file at ~/.cache/nuclei/update-state.gob).
Auth: -auth (configures PDCP_API_KEY via prompt or -auth <key> in v3.2+), -tid team-id, -cup, -aname asset-name.
Interactsh: -iserver oast.live,oast.pro,… (CSV of servers), -itoken (self-hosted), -no-interactsh to disable OAST.
Fuzzing (v3.2+): -ft replace|prefix|postfix|infix, -fm multiple|single, -im jsonl|burp|openapi|swagger|yaml|postman to ingest HTTP traffic from httpx/proxify/Burp.
Uncover bridge: -uc -uq '<query>' -ue shodan,censys to fan out a template over an uncover result set.
Authenticated scans (v3.2+): -sf, -secret-file auth.yaml plus -ps prefetch-secrets. The same secrets file works across all existing templates without per-template edits.

Rate-limit tuning (from the PD optimization guide):

Hardware	`-c`	`-bs`	`-rl`	`-timeout`	`-retries`
1 GB / 1 CPU	2	2	10	30	3
4 GB / 4 CPU	10	10	50	20	2
8 GB / 4–8 CPU	25	25	200	15	1
32 GB+ / 16+	100	100	1000+	10	1

For bug-bounty programs that demand throttling (Intigriti, parts of Bugcrowd), drop to -rl 3 -c 2 and document the values in your report.

Footguns:

"9000 templates" doesn't mean "9000 distinct vulns" — many are variants, language localizations, or detection-only. Use -tags cve,rce + -severity critical,high first; expand if quiet.
Templates with weak matchers cause false positives in waves (issue #12956 was a single bad day in the public set). When a template lights up, verify the match yourself before reporting.
-stop-at-first-match is per-template, not global — and can break workflows that depend on multiple matches.
Default templates include some DoS-shaped ones (fuzz / dos tags) excluded from default scans; only run with -itags if you have permission.
-code is gated behind -enable-self-contained and a signed template by default; that's the safety net for the code protocol introduced in v3.0+.

interactsh — OOB interaction server

interactsh-client -v -server oast.live          # use hosted (default)
interactsh-client -server interact.mydomain.com -token MY-TOKEN -v   # self-hosted

The hosted servers rotate (oast.pro,oast.live,oast.site,oast.online, oast.fun,oast.me). For long-running scans or callbacks containing sensitive metadata, self-host.

Self-host recipe (DigitalOcean / AWS Lightsail / Hetzner — any $5/mo box works):

Buy a domain. In your registrar, add ns1 and ns2 glue records pointing to the VPS IP (Namecheap → Advanced DNS → "Personal DNS Server", or "Add nameserver"). Then change the domain's NS records to ns1.<your-domain>, ns2.<your-domain>.
Open ports on the VPS: 53/udp, 80, 443, 25, 587, 465, optionally 8443 for the web client. LDAP needs 389; SMTPS needs 465.
go install github.com/projectdiscovery/interactsh/cmd/interactsh-server@latest

Run via systemd:

ExecStart=/root/go/bin/interactsh-server \
  -domain interact.mydomain.com \
  -t SOME-RANDOM-TOKEN \
  -cidl 5 -cidn 6 \
  -hd /opt/interact-payloads/

-cidl/-cidn shorten the correlation-id (the random subdomain payload prefix) for cleaner OOB strings.

From nuclei: -iserver interact.mydomain.com -itoken SOME-RANDOM-TOKEN.

Footguns:

The hosted server's correlation IDs are long (c2*...yyyyyn.oast.pro). Some WAFs and SIEMs flag those as suspicious payloads. Self-hosting with -cidl 5 -cidn 6 produces shorter, less obvious strings.
Don't put your interactsh token in shell history — set it as an env var (INTERACTSH_TOKEN) and pass -itoken $INTERACTSH_TOKEN.

cvemap — CLI over CVE databases

cvemap -auth                       # one-time PDCP setup
cvemap -kev -severity critical -product nginx,apache -l 10
cvemap -cs '>8' -es '>0.5' -poc -re -j | jq -r '.[].id' | nuclei -id -

Key flags:

Filters: -id, -v/-vendor, -p/-product, -eproduct, -s/-severity, -cs/-cvss-score '>7', -es/-epss-score '>0.5', -ep/-epss-percentile, -age '<30', -a/-assignee, -vs/-vstatus {modified|confirmed|rejected|new|unconfirmed}.
Toggles (default true): -kev (CISA Known Exploited Vulnerabilities), -t/-template (has a nuclei template), -poc (has public POC), -h1/-hackerone (reported on H1), -re (remotely exploitable).
Output: -f field1,field2, -fe exclude_field, -lsi list-ids, -l 50, -offset, -j/-json, -q '<text>' free-text search, -epk page-keys for interactive paging.

Footguns:

Free-tier query budget per day on the PDCP CVE API — bulk pulls hit limits. Use -q instead of repeated single-CVE queries.
The -template true filter is the fastest way to narrow to "things I can scan today with nuclei".

Phase 4 — Notify & Plumbing

notify — multi-platform alerting

subfinder -d target.com -silent | notify -bulk -id recon
nuclei -l targets -severity critical -o nuc.txt; notify -data nuc.txt -bulk -provider slack,discord

Key flags:

-bulk — assemble all input into one message (chunked at -cl 4000 by default; some providers chunk lower).
-data <file> — input file (or pipe to stdin).
-p, -provider slack,discord,… — restrict providers.
-id recon,vulns — restrict to provider entries with these ids.
-mf, -msg-format 'Hey {{data}}' — format string with helpers {{datetime}}, {{date}}, {{time}}, {{count}}, {{data}}.
-d, -delay 2 — seconds between sends (avoid rate-limiting on Discord webhooks).
-rl, -rate-limit 1 — alternative rate cap.
-pc, -provider-config notify.yaml — alternate provider config.
-proxy http://… — route through your own proxy for auditability.

Provider config (~/.config/notify/provider-config.yaml):

slack:
  - id: recon
    slack_username: notify-bot
    slack_webhook_url: $SLACK_WEBHOOK   # env-var substitution supported
discord:
  - id: vulns
    discord_username: nuclei
    discord_webhook_url: $DISCORD_WEBHOOK
telegram:
  - id: vulns
    telegram_api_key: $TELEGRAM_BOT_TOKEN
    telegram_chat_id: "-1001234567890"
smtp:
  - id: digest
    smtp_server: smtp.gmail.com
    smtp_port: 587
    use_starttls: true
    smtp_password: $SMTP_PASSWORD
gotify:
  - id: ops
    gotify_host: https://gotify.internal
    gotify_token: $GOTIFY_TOKEN

Supported providers: Slack, Discord, Telegram, Pushover, Email (SMTP), Microsoft Teams, Google Chat, Gotify, Custom Webhook.

Footguns:

Webhook URLs are capability secrets — anyone with the URL can post to your channel. Don't put them in command lines (shell history, ps output, CI logs). Always use env-var substitution.
Discord rate-limits aggressively per webhook; -d 2 -cl 1800 is a safe default for bulk firehose runs.
Telegram bot tokens leak the bot's identity — rotate if you accidentally commit the config file. git-secrets or trufflehog --only-verified saves you here.
-bulk concatenates lines into a single message; for line-per- alert behavior (the default), omit -bulk.

Utilities

mapcidr — CIDR math

mapcidr -cidr 173.0.84.0/24 -sbc 10                  # split into 10-host chunks
cat ips.txt | mapcidr -a                              # aggregate IPs to CIDR
cat ips.txt | mapcidr -aa                             # approximate aggregation
mapcidr -cl 'ips.txt,cidrs.txt' -f4                    # combine lists, filter IPv4

cdncheck — CDN/WAF detection

cat ips.txt | cdncheck -resp -silent       # tag each IP with provider

proxify — Swiss army proxy

proxify -o logs/ -dump-req -dump-resp                  # HTTP :8888 / SOCKS5 :10080
proxify -dm 'api.target.com:127.0.0.1'                  # DNS mapping
proxify -mf 'host == "api.target.com"' -req-mf 'remove_header("X-Auth")'

Key flags:

-http-addr, -socks5-addr — listen addresses.
-o logs/ — JSONL-per-request log dir; -dump-req, -dump-resp.
-dm domain:ip,domain:ip — DNS mapping (override resolution).
-mf <DSL> / -req-mf / -resp-mf — match filter + match-and- replace via DSL.
-cert, -key, -out-cert-path — custom MITM CA for transparent TLS interception.
-elastic-url, -kafka-broker — stream into your SIEM.
-upstream-proxy — chain in front of Burp/Caido.

simplehttpserver — Go take on Python's `http.server`

simplehttpserver -listen 0.0.0.0:8000 -upload -basic-auth user:pass -path /opt/share
simplehttpserver -tls -domain target.example.com -cert cert.pem -key cert.key
simplehttpserver -tcp 0.0.0.0:4444 -tls -rules tls-banner.yaml      # full TCP server

Key flags: -listen, -path, -https / -cert / -key, -basic-auth user:pass -realm "…", -upload -max-file-size 100, -cors, -sandbox, -tcp, -rules <yaml>, -verbose.

Use it for: staging RCE payloads (curl --upload-file shell.bin), serving SSRF callback content, running a quick TLS-banner emulator when you're testing detection rules.

aix — LLM CLI

Pipeline Cookbook

The recipes from the original SKILL.md, plus modern additions.

1. Classic external recon → vuln scan

subfinder -d target.com -silent \
  | dnsx -silent -resp -rl 1000 \
  | httpx -silent -title -tech-detect -status-code -jsonl \
  | nuclei -im jsonl -severity critical,high -t cves/

2. Cloud asset → vuln scan

cloudlist -silent -exclude-private -json \
  | jq -r '.host' \
  | httpx -silent \
  | nuclei -t cves/

3. ASN / network fingerprinting

subfinder -d target.com -silent | httpx -silent -asn -jsonl | jq -c '{host, asn}'

4. JARM TLS clustering (find sibling infra)

subfinder -d target.com -silent | httpx -silent -jarm -jsonl \
  | jq -r 'select(.jarm) | "\(.url)\t\(.jarm)"' | sort -k2 | uniq -f1 -c

5. Favicon hash hunting

subfinder -d target.com -silent | httpx -silent -favicon
# then pivot the hash:
uncover -q 'http.favicon.hash:-1234567890' -e shodan

6. Visual recon

subfinder -d target.com -silent | httpx -silent -screenshot -srd /tmp/shots
# review with: feh /tmp/shots/screenshot/ -F

7. Port-scan-driven probing

sudo naabu -host target.com -top-ports 1000 -exclude-cdn -silent \
  | httpx -silent -ports - \
  | nuclei -t exposures/

8. Real-time alerting layer

nuclei -l targets.txt -severity critical,high -silent \
  | notify -bulk -id vulns

9. SPA crawl (modern, JS-heavy targets)

katana -u https://app.target.com -headless -jc -kf all -d 5 -c 10 -p 3 -jsonl \
  | jq -r '.endpoint' \
  | httpx -silent -mc 200,401,403 \
  | nuclei -t exposures/,vulnerabilities/

10. KEV-filtered nuclei (only run templates for currently-exploited CVEs)

cvemap -kev -template -j -silent | jq -r '.[].id' | sort -u > kev-cves.txt
nuclei -l targets.txt -id "$(paste -sd, kev-cves.txt)" -severity critical,high

11. ASN sweep via uncover

uncover -q 'asn:"AS13335"' -e shodan,censys -uf 'host:port' -ul 500 -silent \
  | httpx -silent -title -tech-detect

12. Bug-bounty-friendly throttled chain (program demands rate limiting)

subfinder -d target.com -rl 5 -silent \
  | dnsx -silent -rl 50 \
  | httpx -silent -rl 5 -rlm 60 \
  | nuclei -rl 3 -c 2 -severity critical,high

13. Cert-pivot horizontal recon

tlsx -u target.com -san -cn -resp-only -silent \
  | sort -u | dnsx -silent -resp \
  | httpx -silent

14. Authenticated nuclei scan (v3.2+)

# auth.yaml:
# static:
#   - type: cookie
#     domains: ["target.com"]
#     cookies:
#       - key: session
#         value: $SESSION_COOKIE
nuclei -l targets.txt -sf auth.yaml -ps -severity high,critical

15. ProjectDiscovery Cloud handoff

httpx -auth                              # one-time
httpx -l subs.txt -silent -pd -aname "target.com-prod"   # upload as a named asset group
nuclei -list targets.txt -cloud-upload   # results appear in the dashboard

Common Footguns (cross-cutting)

What happens	Why it bites	How to avoid
`subfinder -all` returns 0 / very few	No provider keys configured; the free sources alone are sparse	Configure ≥5 free-tier providers; verify with `subfinder -ls`
`httpx -screenshot` fills `/tmp`	One PNG per host, no automatic rotation	`-srd /large/disk`; prune per-host older than N days
naabu against `*.cloudflare-protected` is useless	You're scanning the CDN's IP pool, not the origin	`-exclude-cdn` (and validate with `cdncheck` upstream)
nuclei against shared CDN hosts thrashes neighbors	Rate-limit hits multiple tenants in the same /24	`-rl 3 -c 2` for bug-bounty; pair with `httpx -cdn`
FOFA / Censys "no API key defined"	Wrong config format — they need `email:key` / `id:secret`	Single string is the key; use the `:` separator
`dnsx` saturates resolvers	Default rate is unlimited; default `-r` resolvers throttle hard	Always set `-rl`; bring your own resolvers list
`chaos -d target.com` returns nothing	Missing `PDCP_API_KEY` — chaos requires auth, no free unauth path	Get key from `cloud.projectdiscovery.io`; export env var
Webhook URL in shell history	Anyone reading `~/.bash_history` (or your tmux scrollback) can post	Use env-var substitution in `provider-config.yaml`
katana headless OOMs the box	Hundreds of Chromium tabs spawned in parallel	Cap `-p 5 -c 10` first; scale only after watching `top`
Old `nuclei-templates` produce FPs	Templates updated daily; a stale clone misses fixes	Run `nuclei -update-templates` weekly
"interact.sh callback in client logs"	Hosted interactsh sees your payload metadata	Self-host for sensitive engagements
Self-XSS template fires on benign reflection	Generic word matchers, no protocol gating	Pre-condition templates with method + path + tech signals
Pipeline drops blank lines silently	`subfinder	httpx` swallows lines with no protocol prefix

Provider Auth & Secrets

Tool/source	Config path	Env var(s)	Notes
All PD tools	n/a	`PDCP_API_KEY`	Cloud upload, chaos, cvemap, nuclei `-auth`
subfinder	`~/.config/subfinder/provider-config.yaml`	`SUBFINDER_*` per source	`subfinder -ls` lists all sources
chaos	n/a	`PDCP_API_KEY`	Also accepts `-key XXX`
uncover	`~/.config/uncover/provider-config.yaml`	`SHODAN_API_KEY`, `CENSYS_API_ID`+`CENSYS_API_SECRET`, `FOFA_EMAIL`+`FOFA_KEY`, `QUAKE_TOKEN`, `HUNTER_API_KEY`, `ZOOMEYE_API_KEY`, `NETLAS_API_KEY`, `CRIMINALIP_API_KEY`	Env vars take precedence over YAML
cloudlist	`~/.config/cloudlist/provider-config.yaml`	`AWS_`, `GOOGLE_APPLICATION_CREDENTIALS`, `AZURE_`	AWS supports cross-account `assume_role_arn` + `external_id`
notify	`~/.config/notify/provider-config.yaml`	`SLACK_WEBHOOK`, `DISCORD_WEBHOOK`, `TELEGRAM_BOT_TOKEN`, `SMTP_PASSWORD`, `GOTIFY_TOKEN`	Use `$ENV` substitution inside the YAML
interactsh-server	`~/.config/interactsh-server/config.yaml`	n/a	`-t TOKEN` passed at startup; client passes `-token TOKEN`
nuclei v3.2 auth	per-template `auth.yaml` (passed via `-sf`)	per-template variables	Future: HashiCorp Vault / AWS Secrets Manager / 1Password

Secret discipline:

chmod 0600 ~/.config/*/provider-config.yaml.
Add .config/ and *-config.yaml to ~/.gitignore_global.
Never cat or printenv a config in screenshots or asciinema.
Run trufflehog --only-verified filesystem ~/.config after a major refactor — it catches accidental key duplicates.
Rotate the webhook tokens after any leaks; rotate the PD API key if you publish from a CI runner that other people can read.

Modern Integration Patterns

Nuclei v3 → v3.2 migration cheatsheet

-json → -jsonl (the old -json was a JSON array, not lines).
-disable-update-check / -duc is deprecated in v3.7+; new flag is -force-update-check, and an update-state cache lives at ~/.cache/nuclei/update-state.gob.
New protocols: code: (gated behind -enable-self-contained + signed templates), javascript: (LDAP brute-force, Kerberos, custom checks), multi-protocol.
New template features: flow: (orchestrate multiple protocols inside one template), pre-condition: (skip the template unless DSL matches first), self-contained: true (no host required — use for cloud-config-review templates).
Fuzzing: -ft replace|prefix|postfix|infix|replace-regex, -fm multiple|single, plus -im ingest from Burp/JSONL/OpenAPI/ Swagger/Postman.
Authenticated scans: -sf auth.yaml -ps. Works with every existing template — no template edits required.
Signing: -sign with NUCLEI_SIGNATURE_PRIVATE_KEY; unsigned templates can be blocked with -dut, -disable-unsigned-templates.

ProjectDiscovery Cloud Platform (PDCP)

One API key (PDCP_API_KEY) authenticates all PD tools.
<tool> -auth is a one-time interactive setup; persists the key.
httpx -pd/-dashboard uploads probe results to the cloud dashboard. -aname groups them as named assets.
nuclei -cloud-upload (or -cup) uploads scan results. -tid TEAM-ID targets a specific team.
cvemap requires -auth to query the CVE API at all.
Always set ENABLE_CLOUD_UPLOAD=true only if you genuinely want every run to upload — engagement data leaks otherwise.

PD CLI vs BBOT

Use PD CLI when…	Use BBOT when…
You're hunting actively — composing pipelines in a shell	You want hands-off attack-surface monitoring
You want grep-able JSONL you can diff between runs	You want a unified asset graph + module DAG
You're integrating with proxies / Burp / Caido	You want one tool that does everything end-to-end
You need fine rate-limit control per stage	You want sensible defaults you don't think about
You're writing custom Nuclei templates	You want OSINT + cloud + DNS + web in one binary

They're complementary, not exclusive — most operators run BBOT for continuous discovery and the PD pipeline for targeted hunting.

Burp / Caido handoffs

katana -hl -jsonl → import into Burp via proxify as upstream; Burp's history is now the crawl plus the manual replays.
proxify -mf 'contains(path, "/admin")' filters interesting traffic out of a noisy Burp session; pipe the JSONL into nuclei v3.2 via -im jsonl.
Burp's project file → nuclei -im burp -input file.xml re-runs Burp-captured requests as fuzzing seeds.

JSONL → SIEM / report

All PD tools output a common-ish JSONL schema with url/host/ port/status_code/title/tech/time.
notify -data results.jsonl -bulk -mf '{{datetime}} {{data}}' is the lazy way; proxify -elastic-url … or -kafka-broker … is the production way.
For Splunk/Elastic: use httpx/nuclei -store-resp -srd resp/ to capture full bodies, then ship resp/ as evidence alongside the JSONL index.

Sources

Upstream PD documentation and blog posts that informed each section. URLs grouped by tool so a reader can verify any specific claim.

dnsx: https://github.com/projectdiscovery/dnsx • https://docs.projectdiscovery.io/opensource/dnsx/overview • https://docs.projectdiscovery.io/opensource/dnsx/usage

tlsx: https://projectdiscovery.io/blog/a-hackers-guide-to-ssl-certificates-featuring-tlsx

project-discovery-workflow

이 저장소의 다른 Skills

이 저장소의 다른 Skills

ProjectDiscovery Toolchain — Operator Reference

pdtm — Tool Manager

Phase 1 — Discover (map attack surface)

subfinder — passive subdomain enumeration

naabu — fast port scanner

katana — next-gen crawler

cloudlist — multi-cloud asset enumeration

chaos — PD's internet-wide DNS dataset

uncover — query Shodan/Censys/FOFA/Quake/Hunter/ZoomEye/Netlas/CriminalIP/PublicWWW/HunterHow

alterx — DSL-based subdomain permutation

Phase 2 — Enrich (probe & fingerprint)

httpx — the workhorse

dnsx — fast DNS toolkit

tlsx — TLS / cert intel

Phase 3 — Detect (find vulns)

nuclei — template-driven vulnerability scanner

interactsh — OOB interaction server

cvemap — CLI over CVE databases

Phase 4 — Notify & Plumbing

notify — multi-platform alerting

Utilities

mapcidr — CIDR math

cdncheck — CDN/WAF detection

proxify — Swiss army proxy

simplehttpserver — Go take on Python's http.server

aix — LLM CLI

Pipeline Cookbook

1. Classic external recon → vuln scan

2. Cloud asset → vuln scan

3. ASN / network fingerprinting

4. JARM TLS clustering (find sibling infra)

5. Favicon hash hunting

6. Visual recon

7. Port-scan-driven probing

8. Real-time alerting layer

9. SPA crawl (modern, JS-heavy targets)

10. KEV-filtered nuclei (only run templates for currently-exploited CVEs)

11. ASN sweep via uncover

12. Bug-bounty-friendly throttled chain (program demands rate limiting)

13. Cert-pivot horizontal recon

14. Authenticated nuclei scan (v3.2+)

15. ProjectDiscovery Cloud handoff

Common Footguns (cross-cutting)

Provider Auth & Secrets

Modern Integration Patterns

Nuclei v3 → v3.2 migration cheatsheet

ProjectDiscovery Cloud Platform (PDCP)

PD CLI vs BBOT

Burp / Caido handoffs

JSONL → SIEM / report

Sources

ProjectDiscovery Toolchain — Operator Reference

pdtm — Tool Manager

Phase 1 — Discover (map attack surface)

subfinder — passive subdomain enumeration

naabu — fast port scanner

katana — next-gen crawler

cloudlist — multi-cloud asset enumeration

chaos — PD's internet-wide DNS dataset

uncover — query Shodan/Censys/FOFA/Quake/Hunter/ZoomEye/Netlas/CriminalIP/PublicWWW/HunterHow

alterx — DSL-based subdomain permutation

Phase 2 — Enrich (probe & fingerprint)

httpx — the workhorse

dnsx — fast DNS toolkit

tlsx — TLS / cert intel

Phase 3 — Detect (find vulns)

nuclei — template-driven vulnerability scanner

interactsh — OOB interaction server

cvemap — CLI over CVE databases

Phase 4 — Notify & Plumbing

notify — multi-platform alerting

Utilities

mapcidr — CIDR math

cdncheck — CDN/WAF detection

proxify — Swiss army proxy

simplehttpserver — Go take on Python's http.server

aix — LLM CLI

simplehttpserver — Go take on Python's `http.server`

simplehttpserver — Go take on Python's `http.server`