ark-vulnerability-fixer
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches vulnerability data from the CIRCL CVE database (cve.circl.lu) and clones source code from the official project repository (github.com/mckinsey/agents-at-scale-ark). These are legitimate sources for security remediation tasks.
- [COMMAND_EXECUTION]: Employs standard development and package management tools (git, npm, go, pip, make) to perform repository operations and dependency updates. These commands are necessary for the skill's stated purpose.
- [REMOTE_CODE_EXECUTION]: A piped command sequence (curl | python3 -m json.tool) is used to format API responses. Analysis confirms this is for JSON pretty-printing rather than arbitrary code execution.
- [PROMPT_INJECTION]: The skill ingests data from an external CVE API. Ingestion points: curl API call in SKILL.md. Boundary markers: Absent. Capability inventory: git, npm, go, pip, make, and gh commands in SKILL.md. Sanitization: Absent. The risk of indirect prompt injection is mitigated by the mandatory human-in-the-loop protocol.
- [SAFE]: The skill follows security best practices by requiring explicit user confirmation ('STOP AND WAIT') before making any modifications or submitting pull requests.
Audit Metadata