a2a-security-review
Review agent-to-agent (A2A) / multi-agent-system trust: peer identity and authentication, message integrity, capability-negotiation trust, and delegation-chain…
Review a CI/CD pipeline for supply-chain tampering risk — build isolation, runner/agent trust, secret exposure, mutable dependencies, and poisoned-pipeline (PPE) / unauthorized-workflow paths. Use when hardening GitHub Actions, GitLab CI, Jenkins, or similar against build-system
$ npx -y skills add jassics/awesome-claude-security --skill pipeline-integrity-review --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/pipeline-integrity-reviewContext preview
The summary Claude sees to decide when to auto-load this skill.
Review a CI/CD pipeline for supply-chain tampering risk — build isolation, runner/agent trust, secret exposure, mutable dependencies, and poisoned-pipeline (PPE) / unauthorized-workflow paths. Use when hardening GitHub Actions, GitLab CI, Jenkins, or similar against build-system
name: pipeline-integrity-review description: >- Review a CI/CD pipeline for supply-chain tampering risk — build isolation, runner/agent trust, secret exposure, mutable dependencies, and poisoned-pipeline (PPE) / unauthorized-workflow paths. Use when hardening GitHub Actions, GitLab CI, Jenkins, or similar against build-system compromise.
A CI/CD pipeline that an attacker can't subvert to inject code or steal secrets — mapped to recognized risks (OWASP Top 10 CI/CD Security Risks, SLSA build track).
cause attacker-controlled code to run in a privileged build? `pull_request_target` / auto-run on fork PRs with secrets is the classic hole.
ephemeral vs. persistent; can one job tamper with the next?
not printed in logs; OIDC short-lived creds over long-lived tokens.
actions reviewed; base images and toolchains pinned.
protection and required reviews; who can modify pipeline definitions.
(hand to `artifact-provenance-verification`).
1. Map the pipeline: triggers, jobs, runners, secrets, external actions/images, deploy steps, and trust boundaries between them. 2. Walk each OWASP CI/CD risk and PPE/injection path; flag where untrusted input meets privileged execution or secrets. 3. Check pinning, token scope, runner isolation, and review gates. 4. Recommend fixes prioritized by exploitability (privileged code-exec paths first).
A findings list (risk · pipeline location · exploit path · fix) mapped to OWASP CI/CD risks and the SLSA build track, plus quick wins (pin by digest, scope tokens, isolate fork PRs). Formalize with `security-reporting`.
The build system is production: it has the secrets and ships the code, so compromising CI is often easier and higher-impact than compromising prod. Floating action/image tags are silent supply-chain risk — pin by digest. `pull_request_target` with secrets on fork PRs is the single most common critical CI finding.
A Claude Code plugin marketplace for the full cybersecurity & GenAI-security lifecycle — from recon and threat modeling to detection engineering, GRC, and CISO-level strategy. A pentester knows which OWASP test bends a broken-access-control endpoint.
Repo: jassics/awesome-claude-security
Review agent-to-agent (A2A) / multi-agent-system trust: peer identity and authentication, message integrity, capability-negotiation trust, and delegation-chain…
Test the agent execution harness/runtime itself — LangChain/LangGraph, AutoGen, CrewAI, custom ReAct-style loops, or computer-use/browser-use agents — for…
Assess an autonomous / tool-using AI agent for security end-to-end: tool privileges, autonomy and approval boundaries, excessive agency, memory/state…
Test what an AI agent will actually do without human confirmation, including under injected-goal / prompt-injection scenarios, to validate its autonomy and…
Review the security of MCP (Model Context Protocol) servers/clients an agent uses: server trust tier, tool/resource description and result poisoning,…
Inventory the tools/functions an AI agent can call and audit their privileges, side effects, and approval requirements to find excessive-agency and…