raccoon-audit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands including
grep,git,npm, andcurlfor the purpose of auditing and cleaning codebases.\n- [DATA_EXFILTRATION]: The skill is designed to search for and access sensitive information like API keys and environment variables, and includes instructions for usingcurlto interact with external APIs for secret rotation.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted project files without explicit boundaries, which could allow instructions embedded in the audited code to influence the agent.\n - Ingestion points: Extensive file system searches and code inspection of project files (e.g.,
.ts,.js,.py,.env,.json).\n - Boundary markers: Absent; the skill does not provide delimiters or instructions to differentiate between project content and the agent's logic.\n
- Capability inventory: High; the agent is empowered to run shell commands, modify the file system, and perform network requests.\n
- Sanitization: Absent; the content of files being audited is not sanitized before processing.\n- [EXTERNAL_DOWNLOADS]: Recommends the installation and execution of third-party security tools such as
husky,ts-prune,depcheck, andsecretlintfrom standard package registries.
Audit Metadata