{
  "markdown": "# Skill Safety Auditor Plugin for Claude\n\nA security auditing plugin for Claude Code that reviews Claude Code skills for risks before or after installation.\n\n## Installation\n\n### For Claude Code (Marketplace Plugin)\n\n**Via CLI:**\n```bash\nclaude plugin marketplace add mtthwmllr/my-plugin\nclaude plugin install skill-safety-auditor@skill-safety-auditor\n```\n\n**Or manually via config file:**\n```json\n{\n  \"plugins\": [\n    {\n      \"name\": \"skill-safety-auditor\",\n      \"source\": {\n        \"source\": \"github\",\n        \"repo\": \"mtthwmllr/my-plugin\",\n        \"path\": \"plugins/skill-safety-auditor\"\n      }\n    }\n  ]\n}\n```\n\n## Features\n\n### Skills\n\n#### `skill-safety-auditor:skill-safety-auditor` — Skill Safety Auditor\n\nAudits a Claude Code skill for security risks with three flexible modes.\n\n**What it does:**\n- Fetches and reviews skills before you download them (Mode 1)\n- Reviews a downloaded `.skill` file before installation (Mode 2)\n- Audits an already-installed skill from your local directory (Mode 3)\n- Checks for credential theft, prompt injection, obfuscation, and more\n- Produces a structured report with CRITICAL / WARNING / INFO severity ratings\n- Guides you interactively through clearing each warning\n\n**Trigger phrases:**\n- \"Is this skill safe?\"\n- \"Audit this skill before I install it\"\n- \"Can you check this?\" + any skill URL or install command\n- Proactively offered whenever you mention installing a skill\n\n**Checks performed:**\n\n| Check | Severity | What it catches |\n|---|---|---|\n| A1 — Bash / Shell Access | 🟡 WARNING | Shell command execution |\n| A2 — Write / Edit Access | 🟡 WARNING | File write capabilities |\n| A3 — No allowed-tools | 🟡 WARNING | Undefined tool scope |\n| A4 — Overly Broad Tools | 🟡 WARNING | 5+ tools requested |\n| B1 — Credential Access | 🔴 CRITICAL | API keys, tokens, secrets |\n| B2 — Network Calls in Scripts | 🔴 / 🟡 | Outbound HTTP from scripts |\n| B3 — Obfuscated Content | 🔴 CRITICAL | Encoded/hidden payloads |\n| B4 — System Modifications | 🔴 CRITICAL | Persistent shell/cron changes |\n| B5 — Out-of-dir File Access | 🟡 WARNING | Access outside working dir |\n| B6 — Unverifiable Scripts | 🟡 WARNING | Scripts that couldn't be fetched |\n| C1 — Safety Overrides | 🔴 CRITICAL | Prompt injection attacks |\n| C2 — Fake Permissions | 🔴 CRITICAL | False Anthropic/admin claims |\n| C3 — Concealment Instructions | 🔴 CRITICAL | Instructions to hide actions |\n| C4 — Instruction Density | 🟡 WARNING | Off-topic instruction overload |\n| D1 — Anonymous Source | 🟡 WARNING | No maintainer history |\n| D2 — New Repository | 🟡 WARNING | Repo < 30 days old |\n| D3 — Description Mismatch | 🟡 WARNING | Content doesn't match purpose |\n| D4 — Invalid Frontmatter | 🟡 WARNING | Missing/malformed YAML header |\n\n## License\n\nMIT\n",
  "bytes": 2775,
  "sha": "2ff8bcb0a40f3cf6ad2e3e420f07600a21bd3e66bb2fbc6e35e9024375f12eea",
  "repo_slug": "mtthwmllr/skill-safety-auditor-plugin",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_mtthwmllr_skill_safety_auditor_plugin_sk_1c367f1f/readme"
}