exploit

name

exploit

description

Use when running exploitation phase of a pentest, testing for SQL injection, XSS, SSRF, JWT confusion, deserialization, prototype pollution, HTTP smuggling, cache poisoning, SSTI, XXE, GraphQL, SAML, LFI, file upload bypass, CORS, WebSocket hijacking, IDOR, mass assignment, account takeover chains, 2FA bypass, or any class of web/API vulnerability. Also use when the user says "run exploit", "phase 3", "start exploitation", "test for XSS/SQLi/SSRF", or names any specific vuln class.

Exploitation Phase

Philosophy

Every confirmed finding must have a direct evidence quote from the actual response. Never claim Critical without proving the full kill chain: access -> escalation -> data. Never skip a technique because "it probably won't work" - the manifest decides what runs. WAF presence means bypass first, then exploit - never give up at the first 403.

---

Phase 0: Smart Intake

source ~/.claude/skills/_shared/phase0.sh
source ~/.claude/skills/_shared/signals.sh

p0_init_vars "$1"
p0_state_gate || exit 0

p0_read_relay recon secrets
p0_read_hypotheses
p0_read_memory

TECH=$(jq -r '.intel.technologies[]?' "$SESSION" 2>/dev/null | tr '\n' ',')
# Fallback endpoints
[ -z "$API_SPEC_ENDPOINTS" ] && API_ENDPOINTS=$(jq -r '.intel.endpoints[]?' "$SESSION" 2>/dev/null | head -50)
API_ENDPOINTS="${API_SPEC_ENDPOINTS:-$API_ENDPOINTS}"

echo "=== EXPLOIT SMART INTAKE: $TARGET ==="
echo "State: $STATE | WAF: $WAF | Tech: $TECH"
echo "Top hypothesis: [$TOP_HYPO_PROB%] $TOP_HYPO_LABEL"
echo "Verified creds: $(echo "$VERIFIED_CREDS" | grep -c .)"
echo "JWT tokens: $(echo "$JWT_TOKENS" | grep -c .)"
echo "API endpoints: $(echo "$API_ENDPOINTS" | grep -c .)"
echo "ATW flagged (skip): $ATW_FLAGGED"

Tech stack drives technique priority:

Tech detected	Prioritize first
AWS in stack	SSRF->IMDS->IAM, cloud-audit fork
Node.js + JWT	JWT RS256->HS256, prototype pollution
Django/Python	SSTI Jinja2, Python pickle deserialization
Spring/Java	Deserialization (ysoserial), XXE, JNDI
PHP	LFI->RCE (filter chains), PHP type juggling
.NET/ASP.NET	ViewState deserialization, XXE, SSRF
Redis visible	SSRF->gopher->Redis RCE
GraphQL	Introspection, batching, aliasing IDOR
OAuth/SSO	redirect_uri bypass, state CSRF, JWT kid injection
Nginx+Apache chain	Path confusion (CVE-2025-0108), off-by-slash
ElasticSearch	Unauthenticated query access, script injection
MongoDB	NoSQL operator injection, JS injection

---

Phase 1: Build Execution Manifest

MANIFEST_ITEMS="[]"

[ -n "$VERIFIED_CREDS" ] && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e01","tool":"cred_spray","target":"all_live_hosts","reason":"Verified creds from secrets phase","priority":"MUST","status":"pending"}]')

MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e02","tool":"hypothesis_test","target":"'"$TOP_HYPO_LABEL"'","reason":"Top hypothesis '"$TOP_HYPO_PROB"'%","priority":"MUST","status":"pending"}]')

[ -n "$JWT_TOKENS" ] && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e03","tool":"jwt_attacks","target":"jwt_tokens_found","reason":"JWT tokens from secrets","priority":"MUST","status":"pending"}]')

[ -n "$API_ENDPOINTS" ] && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e04","tool":"arjun+nuclei","target":"api_endpoints","reason":"Discovered API surface","priority":"MUST","status":"pending"}]')

MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e05","tool":"dalfox+sqlmap","target":"param_endpoints","reason":"Reflected param surfaces","priority":"SHOULD","status":"pending"}]')

MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e06","tool":"ssrf_hunt","target":"all_live_hosts","reason":"SSRF on URL/file/webhook params","priority":"MUST","status":"pending"}]')

[ "$WAF" != "none" ] && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e07","tool":"403_bypass","target":"blocked_endpoints","reason":"WAF detected","priority":"MUST","status":"pending"}]')

echo "$TECH" | grep -qi "java\|spring" && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e08","tool":"deserialization_java","target":"serialized_endpoints","reason":"Java tech detected","priority":"MUST","status":"pending"}]')
echo "$TECH" | grep -qi "node\|express" && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e09","tool":"prototype_pollution","target":"json_endpoints","reason":"Node.js detected","priority":"MUST","status":"pending"}]')
echo "$TECH" | grep -qi "php" && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e10","tool":"lfi_php_filter_chain","target":"file_param_endpoints","reason":"PHP detected","priority":"MUST","status":"pending"}]')
echo "$TECH" | grep -qi "graphql" && MANIFEST_ITEMS=$(echo $MANIFEST_ITEMS | jq '. + [{"id":"e11","tool":"graphql_full","target":"graphql_endpoint","reason":"GraphQL detected","priority":"MUST","status":"pending"}]')

p0_manifest_write "exploit" "$MANIFEST_ITEMS"
jq '.scalpel.active_manifest.items[] | "[\(.priority)] \(.id): \(.tool) on \(.target)"' $SESSION

---

Phase 2: Technique Loader

Read ONLY the technique files that match your active manifest items. Do NOT load all files.

Use the Read tool to load: ~/.claude/skills/exploit/tech/<filename>

Manifest ID / trigger	Technique file	Load when
e01 - cred spray	cred-spray.md	VERIFIED_CREDS non-empty
e02 - top hypothesis	See hypothesis table below	Always
e03 - jwt attacks	jwt.md	JWT_TOKENS found
e04 - api surface	idor-mass.md + business-logic.md	API endpoints found
e05 - sqli/xss	sqli.md + xss.md	reflected param surfaces
e06 - ssrf	ssrf.md	any live hosts
e07 - waf bypass	Read /403-bypass/SKILL.md	WAF confirmed
e08 - java deser	deser.md	Java/Spring tech
e09 - prototype poll.	prototype.md	Node.js tech
e10 - lfi/php	lfi-upload.md	PHP tech
e11 - graphql	graphql.md	GraphQL tech

Hypothesis-to-technique file mapping:

Hypothesis label contains	Load technique file(s)
SSRF / metadata / IMDS / cloud	ssrf.md
JWT / algorithm / token	jwt.md
SQL / SQLi	sqli.md
XSS / client-side	xss.md
OAuth / OIDC / SSO / SAML	oauth-saml.md
deserialization / ysoserial	deser.md
SSTI / template	ssti-xxe.md
cache poisoning / smuggling	cache-smuggling.md
race condition / TOCTOU	race-timing.md
IDOR / BOLA / mass assign	idor-mass.md
file upload / SVG / polyglot	lfi-upload.md
LFI / path traversal / php filter	lfi-upload.md
GraphQL	graphql.md
business logic / payment / coupon	business-logic.md
prototype pollution	prototype.md
XS-Leaks / timing oracle	xs-leaks.md
CORS / host header	cors-misc.md
WebSocket	cors-misc.md
2FA / MFA / account takeover	2fa-ato.md
JNDI / Log4Shell	jndi-cve.md
Firebase / NoSQL	nosql.md
LDAP injection	nosql.md
PDF / screenshot / wkhtmltopdf	ssrf.md
supply chain / dependency	read /zerodayhunt/SKILL.md
AD / Kerberoast / domain	read /redteam/SKILL.md

Available technique files in exploit/tech/:

File	Covers
sqli.md	SQL injection - detection, sqlmap, OOB, RCE via xp_cmdshell/COPY, WAF bypass
ssrf.md	SSRF detection, AWS IMDS chain, GCP/Azure/k8s, gopher RCE, filter bypass, PDF/screenshot SSRF
jwt.md	alg:none, RS256->HS256, weak HMAC, kid injection, jku/x5u, embedded JWK
cache-smuggling.md	HTTP request smuggling (CL.TE/TE.CL/H2), web cache poisoning + deception
xss.md	DOM XSS sources/sinks, CSP bypass, mXSS, XSS->ATO chains
ssti-xxe.md	SSTI (Jinja2/Twig/FreeMarker/Velocity/Thymeleaf), XXE (XInclude/SVG/XLSX/SSRF)
deser.md	Java ysoserial, PHP phar+PHPGGC, Python pickle, Ruby Marshal, .NET ViewState/JSON.NET
lfi-upload.md	LFI/path traversal, PHP filter chain RCE, log poisoning, zip slip, file upload bypass
oauth-saml.md	OAuth redirect_uri bypass, state CSRF, SAML XSW, XXE via SAML, ADFS attacks
graphql.md	Introspection bypass, IDOR via aliasing, batch attacks, GQL injection
nosql.md	MongoDB operator injection, Elasticsearch script injection, Redis via SSRF, LDAP injection
prototype.md	Server-side PP (EJS/Pug/Handlebars RCE), client-side DOM gadgets
idor-mass.md	IDOR/BOLA/BFLA, mass assignment, API versioning bypass, excessive data exposure
race-timing.md	Race conditions, TOCTOU, insecure randomness, timing oracle, padding oracle
2fa-ato.md	2FA/MFA bypass, password reset flaws, account takeover chains
cors-misc.md	CORS misconfiguration, open redirect chains, WebSocket hijacking, DOM clobbering
xs-leaks.md	XS-Leaks, CSS injection, timing side channels, CRLF, type juggling, ReDoS
jndi-cve.md	JNDI/Log4Shell, recent CVE chains (Spring4Shell, CVE-2025-0108, CVE-2024-38473)
business-logic.md	Business logic, negative prices, workflow bypass, currency manipulation

---

Phase 3: Core Workflow

1. Prioritize MUST manifest items first.
2. Run each manifest item. After each: p0_mark_done <id> in session.json.
3. For every CONFIRMED finding: write Scalpel Finding Record to report_draft.findings[] immediately.
4. Emit signals: emit_signal VULN_CONFIRMED "<title>" "main/exploit" 0.95
5. Run completion gate before finishing: p0_completion_gate || echo "GATE BLOCKED"

Doom loop guard:

TECHNIQUE="exploit:<name>"
RUNS=$(jq -r --arg t "$TECHNIQUE" '.scalpel.doom_loop.technique_runs[$t] // 0' $SESSION)
RUNS=$((RUNS+1))
jq --arg t "$TECHNIQUE" --argjson r $RUNS '.scalpel.doom_loop.technique_runs[$t] = $r' $SESSION > /tmp/s.json && mv /tmp/s.json $SESSION
[ $RUNS -gt 3 ] && echo "[DOOM LOOP] $TECHNIQUE attempted $RUNS times - routing to next" && exit 0

Phase-End - Intel Relay write:

SSRF_JSON=$(jq -r '.intel.internal_ips[]?' $SESSION 2>/dev/null | jq -R . | jq -s . || echo "[]")
CONFIRMED_JSON=$(jq '[.report_draft.findings[] | select(.status=="confirmed") | .title]' $SESSION 2>/dev/null || echo "[]")
AUTH_BYPASS=$(jq 'any(.report_draft.findings[]; .status=="confirmed" and (.title | test("auth|bypass";"i")))' $SESSION 2>/dev/null || echo false)

jq --argjson ssrf "[]" --argjson vulns "$CONFIRMED_JSON" --argjson bypass "$AUTH_BYPASS" \
   '.intel_relay.from_exploit = {
     "ssrf_vectors": $ssrf,
     "internal_ips": [.intel.internal_ips[]? // empty],
     "confirmed_vulns": $vulns,
     "verified_auth_bypass": $bypass
   }' $SESSION > /tmp/s.json && mv /tmp/s.json $SESSION

Tell user: "Exploitation phase complete. interesting_exploit.md written. Run /triage <target> to aggregate findings."

---

Scalpel Finding Record Template

{
  "scl_id": null,
  "title": "<VulnClass> on <Endpoint> leads to <Impact>",
  "date": "<YYYY-MM-DD>",
  "severity": "Critical|High|Medium|Low",
  "confidence": 95,
  "status": "confirmed",
  "technique": "<technique_name_for_ATW>",
  "dna": {"hash": null, "vuln_class": "<class>", "tech_fingerprint": [], "chain_fingerprint": ""},
  "kccg": {
    "score": null,
    "initial_access": {"pass": true, "proof": "<endpoint + method>"},
    "escalation": {"pass": true, "proof": "<privilege/credential gained>"},
    "data_access": {"pass": true, "proof": "<actual data accessed>"},
    "reproducible": {"pass": true, "proof": "3/3 replays identical"},
    "scope_valid": {"pass": true, "proof": "<in-scope confirmation>"},
    "severity_cap": null
  },
  "kill_chain": [
    {"step": 1, "technique": "", "action": "<exact curl command>", "response_fragment": "<quoted response snippet>", "signal": "VULN_CONFIRMED"}
  ],
  "evidence": {"http_proof": "http-responses/<slug>.txt", "supporting_signals": [], "pocs_dir": "pocs/"},
  "impact": {"description": "<business impact>", "cvss_estimate": 0.0},
  "engagement": {"hypothesis": "H1", "skills_chain": ["exploit"], "signals_path": ["VULN_CONFIRMED"]},
  "confirmed_at": "<YYYY-MM-DD HH:MM>"
}