zerodayhunt

name

zerodayhunt

description

Use when hunting for zero-days, backdoors, RCE, supply chain attacks, JWT vulnerabilities, cache poisoning, HTTP smuggling, dependency confusion, source map exposure, GSRM/WAF bypass, internal admin panel exposure, business logic flaws, race conditions, subdomain takeover, cloud misconfigs, mobile APK secrets, OAuth attacks, CORS misconfigs, serialization/SSTI/XXE, CI/CD pipeline attacks, or chained attack vectors. Also use when user says "find zero day", "hunt backdoor", "find RCE", "go deep", "maximum potential", "find critical", "chain attack", or "elite hunt".

Zero-Day & Critical Vulnerability Hunt

Philosophy

Automated tools find known CVEs. This skill finds what scanners miss: logic flaws, architecture leaks, supply chain vectors, auth bypasses, and chained attacks. Evidence required before claiming ANY finding. Never hallucinate - if you didn't see it in a response body, it didn't happen.

The rule that separates elite from script kiddie: A single finding is rarely Critical. The payout comes from CHAINING. SSRF + AWS metadata = IAM creds = cloud takeover. Info disclosure + IDOR = mass PII = Critical. Always ask: "What can I do WITH this?"

Critical rule: NEVER stop at "SSRF/network reachability possible" or "endpoint exists". Must prove full chain: credentials extracted, PII accessed, or data exfiltrated. A 200 with empty body = WAF catch-all, not a real finding.

ARGUMENTS

<target> - domain (e.g. target.com) <focus> - optional: RCE / SSRF / IDOR / SUPPLY-CHAIN / JWT / BUSINESS-LOGIC / CHAIN / ALL

---

Phase 0: Smart Intake

source ~/.claude/skills/_shared/phase0.sh
source ~/.claude/skills/_shared/signals.sh
p0_init_vars "$1"
p0_state_gate "HARVEST" || exit 0
p0_read_relay recon secrets exploit
p0_read_memory
p0_read_hypotheses

# Aliases for downstream phase references
TECH_STACK=$KNOWN_TECH
ALL_ENDPOINTS="$INTERESTING_ENDPOINTS $WAYBACK_ENDPOINTS $API_SPEC_ENDPOINTS"
ATW_FLAGGED=$(echo "$ATW_FLAGGED" 2>/dev/null)

echo "=== ZERODAYHUNT SMART INTAKE: $TARGET ==="
echo "State: $STATE | WAF: $WAF | Tech: $TECH_STACK"
echo "Top hypothesis: $TOP_HYPO_LABEL [$TOP_HYPO_PROB%]"
echo "SSRF vectors from exploit: $(echo "$SSRF_VECTORS" | wc -l)"
echo "Internal IPs available: $(echo "$INTERNAL_IPS" | wc -w)"
echo "JWT tokens: $(echo "$JWT_TOKENS" | wc -l)"
echo "Prior confirmed findings: $(echo "$CONFIRMED_VULNS" | wc -l)"
echo "ATW flagged (skip): $ATW_FLAGGED"
echo ""
echo "Phase selection - run phases where we have targets:"
[ -n "$SSRF_VECTORS" ] && echo "  Phase 8 (SSRF): YES - $(echo "$SSRF_VECTORS" | wc -l) vectors to chain"
[ -n "$JWT_TOKENS" ] && echo "  Phase 7 (JWT): YES - tokens found"
[ -n "$INTERNAL_IPS" ] && echo "  Phase 8 (SSRF pivot): YES - internal IPs to target"
[ -n "$CONFIRMED_VULNS" ] && echo "  Phase 23 (Chain): YES - $(echo "$CONFIRMED_VULNS" | wc -l) confirmed to chain"

Chain-first strategy: Look at what's already confirmed in prior phases and ask "What can I chain WITH this?" before hunting new vectors. SSRF already confirmed + AWS detected = immediately go to IMDSv1 chain. Auth bypass confirmed = immediately check IDOR chain.

Phase 1 - Full Context Build

• SSRF_VECTORS -> target these for cloud metadata chains
• INTERNAL_IPS -> SSRF target list for lateral movement
• JWT_TOKENS -> JWT confusion candidates (check alg header)
• TECH_STACK -> picks specific phases to prioritize (see table below)
• VERIFIED_CREDS -> credential spray on all surfaces before hunting new vulns
• WAF -> Phase 2 first if WAF present
• ALL_ENDPOINTS -> full attack surface for every subsequent phase

Phase priority by tech stack:

Tech detected	Run these phases FIRST
AWS + SSRF_VECTOR	Phase 8 -> Phase 12 (cloud asset chain)
Node.js + JWT	Phase 7 (JWT confusion), Phase 25 (prototype pollution)
Spring Boot	Phase 15 (SSTI/deserialization), Phase 16 (CI/CD)
OAuth/OIDC	Phase 7 (JWT), Phase 31 (SAML/SSO)
GraphQL	Phase 17 (introspection, batching)
React/webpack	Phase 18 (source maps), Phase 5 (JS bundles)
any	Phase 23 (chain blueprints) - always run this last

---

Execution Manifest

Write to session.json.scalpel.active_manifest before starting:

{
  "skill": "zerodayhunt",
  "target": "<target>",
  "items": [
    {"id": "zdh01", "priority": "MUST",     "tool": "waf-header-mining",  "target": "<target>", "status": "pending"},
    {"id": "zdh02", "priority": "MUST",     "tool": "github-js-sourcemap","target": "<target>", "status": "pending"},
    {"id": "zdh03", "priority": "MUST",     "tool": "ssrf-oob",           "target": "<target>", "status": "pending"},
    {"id": "zdh04", "priority": "MUST",     "tool": "jwt-saml-sso",       "target": "<target>", "status": "pending"},
    {"id": "zdh05", "priority": "MUST",     "tool": "chain-blueprints",   "target": "<target>", "status": "pending"},
    {"id": "zdh06", "priority": "SHOULD",   "tool": "business-logic",     "target": "<target>", "status": "pending"},
    {"id": "zdh07", "priority": "SHOULD",   "tool": "supply-chain",       "target": "<target>", "status": "pending"},
    {"id": "zdh08", "priority": "SHOULD",   "tool": "admin-infra",        "target": "<target>", "status": "pending"},
    {"id": "zdh09", "priority": "SHOULD",   "tool": "ssti-deser-xxe",     "target": "<target>", "status": "pending"},
    {"id": "zdh10", "priority": "SHOULD",   "tool": "cors-host",          "target": "<target>", "status": "pending"},
    {"id": "zdh11", "priority": "SHOULD",   "tool": "race-timing",        "target": "<target>", "status": "pending"},
    {"id": "zdh12", "priority": "SHOULD",   "tool": "cicd",               "target": "<target>", "status": "pending"},
    {"id": "zdh13", "priority": "SHOULD",   "tool": "graphql",            "target": "<target>", "status": "pending"},
    {"id": "zdh14", "priority": "IF_TIME",  "tool": "takeover-cloud",     "target": "<target>", "status": "pending"},
    {"id": "zdh15", "priority": "IF_TIME",  "tool": "mobile-apk",         "target": "<target>", "status": "pending"},
    {"id": "zdh16", "priority": "IF_TIME",  "tool": "client-proto",       "target": "<target>", "status": "pending"},
    {"id": "zdh17", "priority": "IF_TIME",  "tool": "file-crypto",        "target": "<target>", "status": "pending"},
    {"id": "zdh18", "priority": "IF_TIME",  "tool": "websocket-api",      "target": "<target>", "status": "pending"},
    {"id": "zdh19", "priority": "IF_TIME",  "tool": "xs-leaks",           "target": "<target>", "status": "pending"},
    {"id": "zdh20", "priority": "MUST",     "tool": "output",             "target": "<target>", "status": "pending"}
  ]
}

---

Technique Loader

Load the relevant technique file from ~/.claude/skills/zerodayhunt/tech/ based on active manifest item:

Manifest item	File	Phases covered
waf-header-mining	tech/waf-header-mining.md	Phase 2, 3, 20
github-js-sourcemap	tech/github-js-sourcemap.md	Phase 4, 5, 18
supply-chain	tech/supply-chain.md	Phase 6
jwt-saml-sso	tech/jwt-saml-sso.md	Phase 7, 31
ssrf-oob	tech/ssrf-oob.md	Phase 8, 32
business-logic	tech/business-logic.md	Phase 9, 27
race-timing	tech/race-timing.md	Phase 10, 21
takeover-cloud	tech/takeover-cloud.md	Phase 11, 12
mobile-apk	tech/mobile-apk.md	Phase 13
cors-host	tech/cors-host.md	Phase 14
ssti-deser-xxe	tech/ssti-deser-xxe.md	Phase 15
cicd	tech/cicd.md	Phase 16
graphql	tech/graphql.md	Phase 17
admin-infra	tech/admin-infra.md	Phase 19, 28
websocket-api	tech/websocket-api.md	Phase 22
chain-blueprints	tech/chain-blueprints.md	Phase 23
client-proto	tech/client-proto.md	Phase 24, 25
file-crypto	tech/file-crypto.md	Phase 26, 29
xs-leaks	tech/xs-leaks.md	Phase 30
output	tech/output.md	Evidence Classification + Output + Phase-End

---

Hypothesis-Driven Prioritization

After Phase 0, reprioritize manifest based on intel:

Condition	Move to MUST
AWS_HINT=true and any SSRF vector	ssrf-oob, chain-blueprints (Chain B/H)
JWT_TOKENS found	jwt-saml-sso
TECH_STACK contains Node.js	client-proto (prototype pollution)
TECH_STACK contains Java/Spring	ssti-deser-xxe
GraphQL endpoint discovered	graphql
GitHub org found in recon	github-js-sourcemap, supply-chain, cicd
E-commerce target	business-logic, race-timing
SSRF_VECTORS from exploit relay	ssrf-oob (escalate existing)
CONFIRMED_VULNS has any finding	chain-blueprints (always)

Skip ATW-flagged techniques. Doom loop guard in output.md.

---

Signal Reference

All signals emitted via emit_signal from _shared/signals.sh:

Signal type	When	Confidence
VULN_CONFIRMED	Full chain proven with evidence	0.93-0.97
SURFACE_FOUND	New attack surface (takeover, supply chain)	0.88-0.90
CRED_FOUND	Credential discovered and verified	0.92
POTENTIAL	Probe positive but chain unconfirmed	0.60-0.75