murrtada/bug-bounty-agent-skills / skills/bb-recon
bb-recon
Verifiedbb-recon — Discovery Methodology & Tool Commands
Scope: passive + low-noise active recon against in-scope web targets. Load this skill
in recon-executor and content-collector tasks. Every command below must pass the
scope guard first (python3 bb-ops/lib/scope.py check <target> <host>).
0. Pre-flight
- Read
memory/targets/<target>/target.md: scope, creds, OAST callback, auth status. - Record OAST callback for SSRF/interaction detection.
1. Subdomain enumeration (passive first, then active)
subfinder -d example.com -all -o subs.txt
amass enum -passive -d example.com
assetfinder --subs-only example.com
curl -s "https://crt.sh/?q=%25.example.com&output=json" | jq -r '.[].name_value' | sort -u
curl -s "https://api.certspotter.com/v1/issuances?domain=example.com&include_subdomains=true&expand=dns_names" | jq -r '.[].dns_names[]' | sort -u
# resolve + de-dup
cat subs.txt | sort -u | dnsx -silent -a -resp -o resolved.txt
# live hosts + tech
cat resolved.txt | httpx -silent -title -server -status-code -tech-detect -o live.txt
2. URL / JS / historical collection
katana -u https://app.example.com -d 3 -jc -o urls.txt # crawl + JS analysis
gau --subs example.com | sort -u > historical.txt # wayback + commoncrawl + otx
waybackurls example.com | sort -u >> historical.txt
# JS bundle analysis: download all script tags, grep for endpoints
grep -rhoE "https?://[a-zA-Z0-9./_-]+" bundle.js | sort -u
grep -rhoE '"/[a-zA-Z0-9_/-]+"' bundle.js | sort -u # API path candidates
grep -rniE "secret|token|apikey|api_key|password|client_id" bundle.js
3. Technology fingerprinting
httpx -l live.txt -silent -tech-detect -json -o tech.json
curl -sI https://host | sed -n '1p;/server/i' # manual header check
# detect known-framework endpoints
# Express: / -> x-powered-by
# Spring: /actuator, /swagger-ui.html, /v3/api-docs
# Django: /admin/
# GraphQL: /graphql, /query, /gql (POST, content-type application/json)
# Keycloak: /auth/realms/{realm}/.well-known/openid-configuration
4. GitHub / secret recon (in-scope orgs/repos only)
gh search repos --owner <org> --limit 100
gh api "orgs/<org>/repos?per_page=100" | jq -r '.[].clone_url' | while read u; do git clone -q "$u" /tmp/recon/$u; done
gitleaks detect --source /tmp/recon --report-format=json --report-path=gitleaks.json
trufflehog git https://github.com/<org>/<repo> --json
grep -rniE "(api[_-]?key|client[_-]?secret|BEGIN (RSA|OPENSSH|EC) PRIVATE|password *= *[^ ]+)" /tmp/recon
5. Live probe + port discovery (web only)
naabu -l live.txt -top-ports 1000 -o ports.txt # web-adjacent ports
nuclei -l live.txt -t http/exposures -t http/misconfiguration -severity low,medium,high
6. Parameter discovery (after endpoints exist)
arjun -u https://host/api/endpoint -m GET,POST --stable
x8 -u https://host/api/endpoint --wordlist /usr/share/seclists/Discovery/Web-Content/burp-parameter-names.txt
7. LLM-assisted source research (multi-agent prompt)
Adam Kues' approach (Searchlight Cyber, 2026): a long-running prompt designed for aggressive multi-agent use, scoped to decompiled/downloaded source, can autonomously find full chains. Reusable pattern for our targets:
1. DOWNLOAD source first (SPA bundles, sourcemaps, decompiled jars/classes when available).
2. Scope the prompt tightly: <TARGET + exact component>, "find a pre-auth RCE" style objective.
3. Give the agent the artifact paths and read access; require it to PROVE each step with the
actual code/endpoint before proposing it.
4. Instruct it to emit a numbered chain (each step independently testable), matching our
chain-link discipline (verify.dag.yaml / report.dag.yaml gates).
5. Human verifies + reproduces every step non-intrusively. Never submit a model-proposed
step that wasn't reproduced.
- Lesson source: JFrog Artifactory pre-auth RCE (edrabb.fr) found via a Codex run pointed at decompiled source using this prompt style — trailing-slash bypass -> refresh token -> claim confusion -> Logback RCE.
- Always keep scope guard + program rules (no scanning/DoS) in the prompt too.
Output discipline
- Every result →
lib/dedup.py upsert <target> asset|endpoint <record>. - Raw artifacts →
memory/targets/<target>/evidence/. - Emit one
evidence.recordedevent per meaningful batch (1-line summary). - Never paste full tool dumps into your reply — summarize.
Safety
- Passive = default. No active fuzzing against hosts without an in-scope confirmation.
- If a subdomain resolves to an IP on a third-party CDN (e.g. S3/Cloudfront without the owner's DNS), stop and flag for scope review — don't scan.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 197k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |