offensive-osint – Security skill for AI agents

Bundled upstream support

Supporting files are included at skill-relative paths. See [UPSTREAM-NOTICE.md](UPSTREAM-NOTICE.md) for their pinned sources, licenses, hashes, and import changes.

> ⚠️ AUTHORIZED USE ONLY

> This skill is for educational purposes or authorized security assessments only.

> You must have explicit, written permission from the system owner before using this tool.

> Misuse of this tool is illegal and strictly prohibited.

> Mandatory confirmation gate

> Before running any command that probes, exploits, changes, persists on, extracts data from, or attempts credential access against a target:

> 1. Ask the user to state the exact target URL, IP, account, or resource.

> 2. Ask the user to confirm written authorization and the permitted scope.

> 3. Show the exact command(s) and explain their expected effect.

> 4. Wait for explicit confirmation in the current conversation.

>

> Without that confirmation, remain read-only and provide defensive guidance only. Prefer a sandbox, disposable VM, or controlled lab.

Offensive OSINT — External Red-Team Arsenal

> v3.0 — Refactored 2026-05-02 from a 4,168-line monolith into a lean SKILL.md (~400 lines) plus 15 modular reference files in references/. Detail content loads on demand — Claude reads only the reference files relevant to the current task.

0. When to use / When NOT

Use this skill when:

• You need concrete probe paths, wordlists, regexes, payloads, scoring rules, or tool URLs.
• You're executing reconnaissance and need the actual technical reference (vs. methodology).
• You're building a recon automation and need specific lists to seed it.

Do NOT use this skill when:

• The user is asking for active exploitation, post-exploitation, or anything past reconnaissance.
• The user is asking for defensive / blue-team detections.
• The target's authorization isn't established — see §1.

---

1. Authorization & Legal Posture

For assets the operator owns or has written authorization to assess. The mandatory confirmation gate above applies to every active request and credential use. Modes, scores, discovered assets and companion skills never grant authorization. Keep passive source research distinct from requests to target systems; obtain approval before sending private target identifiers to a third-party provider.

---

2. Confidence Levels

• TENTATIVE — plausible based on indirect evidence (snippet-only dork match, single-source asset, inferred email pattern).
• FIRM — directly observed (subdomain resolves, HEAD-confirmed bucket exists, banner returned).
• CONFIRMED — a precisely stated observation verified by independent corroboration or approved direct evidence. Credential identity acceptance, multiple agreeing sources and one object read do not prove unrestricted permissions, independent sourcing or unintended disclosure.

---

3. Output Format Conventions

Findings should carry: id, module, asset_key, category, severity (info/low/medium/high/critical), confidence, title, description, evidence (url + UTC timestamp + sha256 + redacted excerpt ≤ 2 KiB; protected raw evidence stored separately), references, remediation. UTC timestamps everywhere.

---

4. Source Hygiene & Citations

URL + UTC timestamp + SHA-256 + tool version + run_id, every artifact. PNG screenshots, JSONL run logs, redacted report excerpts capped at 2 KiB body. Collect only the minimum approved evidence; raw captures belong in a private engagement directory with restrictive permissions, retention and access controls.

---

5. Do NOT

• Don't paste creds/PII/session tokens into cloud LLMs.
• Do not run destructive probes in this reconnaissance skill. DEEP/--aggressive is not authorization; exploitation, writes, persistence and credential cracking require a separately approved assessment or controlled lab.
• Don't use validated credentials for anything except read-only liveness check.
• Don't single-source attribute.
• Don't assume vendor labels are ground truth.

---

6. General OSINT (curated tool refs)

• [OSINT Bookmarks](https://tools.myosint.training/) — comprehensive bookmarks.
• [OSINT Framework](https://osintframework.com/) — tool/resource directory.
• [IntelTechniques Tools](https://inteltechniques.com/tools/) — investigative suite.
• [Bellingcat Toolkit](https://www.bellingcat.com/resources/2024/09/24/bellingcat-online-investigations-toolkit/) — investigative journalism.
• [CyberSudo OSINT Toolkit](https://docs.google.com/spreadsheets/d/1EC0sKA_W9znzsxUt0wye9UYtyATXw5m8) — OSINT websites list.
• [Google Dorks](https://dorksearch.com/) — efficient Google searching.
• [Distributed Denial of Secrets](https://ddosecrets.com/) — leaked datasets.
• [Country-Specific Resources](https://digitaldigging.org/osint/) — country-targeted OSINT.

---

How to use this skill

This skill is a lean operational index. Most concrete data (wordlists, regexes, dorks, endpoint catalogs, severity examples) lives in the references/ subfolder, organized by topic.

Workflow when this skill triggers:

1. Read this SKILL.md to anchor on principles (§0-5), scoring rubrics (§20-21), attack-path templates (§39), and the references index below.

2. For task-specific data, read only the reference file(s) you need — do NOT pull all 15. Each reference is self-contained.

3. Use the bug-bounty skill for the local toolkit at ~/security-research/bug-bounty-resources/ and osint-methodology for the planning framework.

Loading rules of thumb:

• Single-class question (e.g., "what's the regex for AWS keys?") → load secret-patterns.md only.
• Multi-class engagement (e.g., "do an external recon on target.com") → load probes-and-wordlists.md first, then add others as the engagement narrows.
• Severity / triage question → load severity-matrix.md.

---

References Index

| File | Coverage | Trigger phrases |

|---|---|---|

| [probes-and-wordlists.md](references/probes-and-wordlists.md) | API/Swagger/GraphQL paths, cloud-bucket arsenal, JS guess-paths, vendor & cloud-native fingerprints, K8s/CI-CD exposure, doc/wiki leaks, WHOIS/RDAP, DNS catalog, Wayback CDX, copy-paste curl probes, email security analysis, origin/CDN bypass | swagger discovery, graphql introspection, subdomain takeover, cloud bucket enum, S3/GCS/Azure enum, kubernetes exposure, CI CD exposure, vendor fingerprint, WHOIS RDAP, Wayback CDX, copy paste probes, curl one-liner |

| [identity-fabric.md](references/identity-fabric.md) | Concrete endpoints for Entra/Okta/ADFS/Google/SAML, M365 deep (Teams federation, SharePoint, OneDrive), GraphQL field-suggestion enumeration, user-enum patterns | identity fabric, SSO discovery, IdP fingerprinting, okta enum, entra enum, azure AD enum, ADFS enum, SAML metadata, Microsoft 365 deep, Teams federation, SharePoint enum, OneDrive enum, graphql field suggestion |

| [secret-patterns.md](references/secret-patterns.md) | 48-pattern secret-regex catalog (AWS, GCP, GitHub PATs, Stripe, Slack, JWT, private keys, Anthropic/OpenAI/HuggingFace, Cloudflare, DigitalOcean, npm, PyPI, Docker Hub, Atlassian, DataDog, Sentry, ngrok) with severity & FP notes | secret scanning, secret leak, leaked credential, JWT triage, AWS key triage, Anthropic API key, OpenAI API key |

| [secret-validators.md](references/secret-validators.md) | 9 read-only secret validators + post-discovery enumeration workflows for AWS/GitHub/Slack/Postman/JWT/Anthropic/OpenAI/npm/Atlassian/DataDog | secret validation, post discovery workflow, AWS key triage, JWT triage |

| [dork-corpus.md](references/dork-corpus.md) | 80+ Google/Bing/DDG dork templates across 9 categories + 13 GitHub code-search dorks tailored for targets | google dorking, bing dorking, github dorking, dork corpus |

| [recon-stack.md](references/recon-stack.md) | Subdomain-source stack (passive & active), infrastructure & attack-surface OSINT (Shodan/Censys/crt.sh/JARM/favicon mmh3), TLS deep audit, reverse DNS, IPv6 enumeration | subdomain enumeration, certificate transparency, crt.sh, shodan recon, censys recon, JARM, favicon mmh3, TLS deep audit, JA3 JA4, reverse DNS sweep, IPv6 enumeration |

| [breach-and-credentials.md](references/breach-and-credentials.md) | Breach & leak data sources (HudsonRock, HIBP, DeHashed, IntelX, infostealer logs), email-pattern inference, email-harvest source stack | breach lookup, have I been pwned, HudsonRock cavalier, infostealer, dehashed, intelx, email harvest |

| [people-osint.md](references/people-osint.md) | Search engines, username & email investigation, people search, phone OSINT, social media, public records & company info | username investigation, people search, phone OSINT, social media OSINT, public records |

| [saas-public-surfaces.md](references/saas-public-surfaces.md) | Postman public workspace search (verified endpoint), Stack Exchange OSINT sweep, public SaaS dork stack (Notion, Confluence, Trello) | postman workspace, stack exchange OSINT, Notion p

… (full content is in the download package)