/safety-review
Run an AI safety review for a feature/model — harms, evaluations, guardrails, and a documented safety case.
$ npx -y skills add jassics/awesome-claude-security --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/safety-review
Context preview
What this command does when you run it.
Run an AI safety review for a feature/model — harms, evaluations, guardrails, and a documented safety case.
Command definition
safety-review.mddescription: Run an AI safety review for a feature/model — harms, evaluations, guardrails, and a documented safety case. argument-hint: [AI feature / model / release to review]
Run an AI safety review for: **$ARGUMENTS**
Walk it, using installed skills (note any whose plugin is missing). This is **safety** (harm, bias, misuse), distinct from security:
1. **Harm modeling** — `/ai-safety:harm-modeling` to enumerate plausible harms, affected groups, and misuse paths. 2. **Evaluate** — `/ai-safety:safety-evaluation` and `/ai-safety:safety-red-team` to test for the modeled harms; `/ai-safety:bias-fairness-assessment` for disparate impact. 3. **Guardrails** — `/ai-safety:guardrail-review` to check the mitigations actually cover the evaluated risks. 4. **Safety case** — `/ai-safety-engineer:safety-case` to assemble the structured, evidence-backed argument that the system is acceptably safe to ship. 5. **Report** — `/security-reporting:executive-summary` for the go/no-go; `/security-diagramming:mindmap` for the harm/mitigation map.
For deep execution, hand off to the `ai-safety-engineer` agent. The safety case is the deliverable — claims must be backed by evaluation evidence, not assertions.
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
Other commands on awesome-claude-security.
- /defend
Run a threat-informed defense cycle for a technique or threat — coverage check, hunt, detection, and purple-team validation.
Open command - /board-brief
Build a board/executive security brief — strategy, quantified risk, and a board-ready deck.
Open command - /tech-review
Assess technology/security risk for a strategic decision and frame the secure-by-design path.
Open command - /precommit
Run the pre-commit security gate on the current changeset and report a single pass/fail verdict.
Open command - /assessment
Run a compliance gap-assessment for a framework, tie gaps to risk, and produce findings + remediation.
Open command - /engagement
Run an authorized penetration test end-to-end, chaining recon, testing, and reporting skills into one flow.
Open command