@aegis-scan/skills

Opt-in skill library for Claude Code and compatible AI agents. Third sibling in the AEGIS full-repertoire institutional-grade security toolkit.

AEGIS is a three-layer toolkit

• @aegis-wizard/cli — scaffold + agent-brief generator. Builds a secure Next.js + Supabase + shadcn SaaS from day one with an agent-consumable Markdown brief.
• @aegis-scan/cli — defensive SAST scanner (five-package family). Catches what the scanner knows to look for across your built application.
• @aegis-scan/skills — this package. Red-team methodology library (v0.1.0) with defensive, MITRE-mapped, and ops extensions landing in later releases. Primes your AI coding-agent with attack-class decision-trees so you can stress-test what you built before shipping.

Build with the wizard. Scan what you built. Test it red-team-style. Full lifecycle, one toolchain, one attribution-compliant open-source license stack.

Quickstart

npm install -g @aegis-scan/skills

# Install every skill into Claude Code's user-skill directory
aegis-skills install

# List what is available
aegis-skills list

# Inspect a specific skill
aegis-skills info sqli

After install lands the skill files under ~/.claude/skills/user/aegis-skills/, Claude Code auto-loads each SKILL.md based on its trigger-phrases whenever you invoke the agent with a relevant prompt.

What ships

Offensive skills — skills/offensive/snailsploit-fork/

Thirty-seven offensive-security SKILL.md files covering:

• Web application: sqli · xss · ssrf · ssti · xxe · idor · file-upload · rce · deserialization · race-condition · request-smuggling · open-redirect · parameter-pollution · graphql · waf-bypass (15)
• Auth and identity: jwt · oauth (2)
• Infrastructure and binary: shellcode · edr-evasion · exploit-development · exploit-dev-course · basic-exploitation · crash-analysis · mitigations · windows-mitigations · windows-boundaries · keylogger-arch · initial-access · advanced-redteam (12)
• Reconnaissance and OSINT: osint · osint-methodology (2)
• Fuzzing and vulnerability research: fuzzing · fuzzing-course · bug-identification · vuln-classes (4)
• AI security: ai-security (1)
• Utility: fast-checking (1)

All forked from SnailSploit/Claude-Red under MIT License with attribution preserved per-file.

Defensive skills — skills/defensive/aegis-native/

Three AEGIS-native SKILL.md files (MIT) mirroring @aegis-wizard/cli patterns and providing remediation guidance for @aegis-scan/cli findings:

• rls-defense — Supabase Row-Level Security hardening (covers rls-bypass-checker + template-sql-checker findings)
• tenant-isolation-defense — multi-tenant SaaS isolation (covers tenant-isolation-checker + mass-assignment-checker findings)
• ssrf-defense — SSRF defense including DNS-rebinding, IPv6, cloud metadata-endpoint protection (covers ssrf-checker + cross-file taint findings)

MITRE-mapped skills — skills/mitre-mapped/aegis-native/

Three AEGIS-native SKILL.md files (MIT) cross-walking AEGIS findings to MITRE frameworks:

• mapping-overview — top-level per-CWE → ATT&CK technique mapping plus tactic-level coverage summary; ATLAS overlay for AI/LLM threats; D3FEND defensive-countermeasure mapping; NIST CSF 2.0 + NIST AI RMF function-level alignment.
• t1190-exploit-public-app — deep-dive on T1190 (the #1 Initial Access vector in Verizon DBIR 2024).
• t1078-valid-accounts — deep-dive on T1078 (Valid Accounts) coverage via the AEGIS credential-protection scanner family.

Operations skills — skills/ops/aegis-native/

Three AEGIS-native SKILL.md files (MIT) wrapping the AEGIS workflow in process-discipline:

• triage-finding — operational runbook for triaging an AEGIS finding (severity → confidence → verify → fix-vs-suppress-vs-defer).
• suppress-correctly — when suppression is appropriate, the structured-rationale syntax, anti-patterns, and audit-trail expectations.
• escalation-runbook — what to do when a BLOCKER reaches main, when a finding suggests active exploitation, or when a credential leak is detected.

Compliance skills — skills/compliance/aegis-native/

One AEGIS-native multi-file SKILL.md (MIT) for adversarial DE/EU compliance audits:

• brutaler-anwalt — adversarial DE/EU compliance auditor (DSGVO / DDG / TTDSG / UWG / NIS2 / EU AI Act / branchenrecht / strafrecht-steuer) with three-persona self-verification (Hunter / Challenger / Synthesizer). Slash-command activation via /anwalt. Ships an 11-file references/ sibling tree (~120 KB) with per-bereich rules, BGH/EuGH-judgment database, abmahn-templates, and an explicit AEGIS-scanner-output → rechtliche-Bewertung mapping. The installer auto-copies the references tree alongside the SKILL.md.

Required external skills (mandatory complement, not forked)

For full Supabase / Postgres coverage, AEGIS users must also install the upstream Supabase skill package — installed once, globally, via the upstream maintainer's distribution channel:

npx skills add supabase/agent-skills -g -y

This installs two MIT-licensed skills from supabase/agent-skills:

• supabase — comprehensive Supabase development skill: Auth / Edge Functions / Realtime / Storage / Vectors, client libraries (supabase-js, @supabase/ssr) integration into Next.js / SvelteKit / Astro / Remix, JWT and session security, Supabase CLI workflow (migration new, db query, advisors), MCP server troubleshooting, schema-change workflow (execute_sql vs apply_migration), Postgres-extension setup (pg_graphql / pg_cron / pg_vector).
• supabase-postgres-best-practices — 8-category Postgres performance guide (30+ reference files): query optimization (covering / partial / composite indexes), connection management (pooling / limits / prepared statements / idle timeout), schema design (primary keys / FK indexes / partitioning / constraints), concurrency & locking (advisory / deadlock prevention / skip-locked), data access (batch inserts / N+1 / pagination / upsert), monitoring (EXPLAIN ANALYZE / pg_stat_statements / vacuum), and advanced features (full-text search / JSONB indexing).

The AEGIS-native defensive skills (rls-defense, tenant-isolation-defense) cross-reference the upstream security-rls-basics.md, security-rls-performance.md, and security-privileges.md reference files for performance-tuned RLS patterns and least-privilege role design — install the upstream package so those cross-references resolve at use-time.

See ATTRIBUTION.md for the full rationale on why these skills are mandate-via-upstream rather than forked into this tree.

Attribution + license

See ATTRIBUTION.md for the full credit chain. Offensive skills are MIT-via-upstream-fork; defensive / mitre-mapped / ops / compliance skills are MIT-AEGIS-original.

Multi-source architecture

@aegis-scan/skills is designed to grow across sources without re-architecting the package. The skills/ tree carries five category-directories:

skills/
├── offensive/
│   └── snailsploit-fork/         — 37 SKILL.md files (MIT, forked from SnailSploit/Claude-Red)
├── defensive/
│   └── aegis-native/             — 3 SKILL.md files (MIT, AEGIS-original)
├── mitre-mapped/
│   └── aegis-native/             — 3 SKILL.md files (MIT, AEGIS-original — ATT&CK / ATLAS / D3FEND / NIST cross-walk)
├── ops/
│   └── aegis-native/             — 3 SKILL.md files (MIT, AEGIS-original — triage / suppress / escalation runbooks)
└── compliance/
    └── aegis-native/             — 1 multi-file SKILL.md + 11-file references/ tree (MIT, AEGIS-original — adversarial DE/EU compliance auditor with three-persona self-verification)

Total: 47 skills across 5 categories and 2 source-namespaces (snailsploit-fork for the offensive fork; aegis-native for the defensive / mitre-mapped / ops / compliance AEGIS-original content).

When future external sources land, they slot into the existing tree under their own per-source subdirectory (e.g., defensive/anthropic-cybersec-pick/) and the manifest metadata expands without layout churn.

Structural invariant

The skills/ directory is markdown-only by construction. No executable content, no binaries, no install-time lifecycle scripts anywhere in the package. The publish-skills.yml CI gate enforces this structurally before every tag-push. A consumer running npm install @aegis-scan/skills executes zero scripts from the @aegis-scan namespace. See the top-level SECURITY.md for the full supply-chain integrity posture.

Responsible use

This package ships offensive-security methodology for authorized use only:

• Authorized security testing of systems you own or have explicit written permission to test.
• Bug-bounty engagements strictly within the defined scope.
• CTF competitions and educational environments.
• Defensive security research — understanding attack classes to defend against them.

Use against systems you do not own or have permission to test is unauthorized and likely illegal. AEGIS provides methodology; responsible use is the operator's obligation. See the top-level SECURITY.md for the full responsible-use disclosure and the SECURITY-INCIDENT-RESPONSE.md for the abuse-report channel.

License

MIT (see LICENSE). Upstream skills ship under their original licenses with attribution preserved per-file — see ATTRIBUTION.md for the full chain.