Research: agentic-actions-auditor
Cached research evidence for agentic-actions-auditor (not authority).
Purpose
Section titled “Purpose”Audits GitHub Actions workflows for AI agent security vulnerabilities specific to integrations like Claude Code Action, Gemini CLI, OpenAI Codex, and GitHub AI Inference. Detects 9 attack vectors (env var intermediary, direct ${{}} injection, gh CLI data fetch at runtime, pull_request_target+checkout, error log injection, subshell bypass, eval of AI output, dangerous sandboxes like —yolo, wildcard allowlists). From Trail of Bits skills marketplace.
Harness Coverage
Section titled “Harness Coverage”Target agents: antigravity, claude-code, codex, crush, cursor, gemini-cli, github-copilot, grok, opencode (per install). Activates automatically on detection of relevant .github/workflows YAML containing AI agent actions; also direct invocation.
Trust And Risks
Section titled “Trust And Risks”trust_tier=needs-inspection; status=inspect-then-install; provenance=verified-install-command; risks=High impact CI context abuse and injection if workflows integrate AI agents with write perms; focus is detection not exploitation. Review triggers, allowed tools, and env flows before trusting in prod repos. policy=Inspect source, hooks, scripts, credentials, and dedupe before install.; evidence=Curated inspect-then-install batch for trailofbits/skills in authoring catalog; live repo https://github.com/trailofbits/skills (5.7k stars, CC-BY-SA-4.0, security research org).
Install Prerequisites
Section titled “Install Prerequisites”Install: npx skills add trailofbits/skills --skill agentic-actions-auditor -y -g -a antigravity claude-code codex crush cursor gemini-cli github-copilot grok opencode status=inspect-then-install; selector=named. Best with .github/ workflows present. Run /review source first per policy.
Upstream Maintainer
Section titled “Upstream Maintainer”trailofbits/skills — security-focused plugin marketplace and skills for Claude Code / agent harnesses. See also related static-analysis, variant-analysis, differential-review plugins.
Comparable Alternatives
Section titled “Comparable Alternatives”differential-review for security PR/diff analysis with history; static-analysis (codeql/semgrep) or insecure-defaults for other vuln classes; general review. A general-purpose agent instruction without a scoped skill contract for GitHub Actions AI agent security audits.
> Web evidence from github.com/trailofbits/skills README + plugin/agentic-actions-auditor/README.md + SKILL structure (fetched 2026-06); cross-ref docs/src/authoring/skills. Evidence only; not an endorsement.