skills / security
amaniagent · category

security

Skills that vet other skills, repos, and agent output before you trust them — all on the same 0–8 danger scale.

012345678

answer-verifier

Audits an agent's answer / output for hallucination & fabrication. Click for the full skill + reports.

dependency-auditor

Audits a project's dependencies / lockfile for supply-chain risk. Click for the full skill + reports.

mcp-auditor

Audits an MCP server's tools, scopes & config. Click for the full skill + reports.

prompt-injection-detector

Audits untrusted content an agent ingests for embedded prompt injection. Click for the full skill + reports.

repo-auditor

Audits a whole GitHub repo / package. Click for the full skill + reports.

settings-auditor

Audits shipped agent config / hooks (.claude/settings.json). Click for the full skill + reports.

skill-auditor

Audits a single agent skill / plugin / MCP. Click for the full skill + reports.

→ Tutorial: how to spot a malicious skill or repo