repo-deep-dive
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands such as
find,git, andwcto analyze codebase statistics and structure (references/phases.md). - [INDIRECT_PROMPT_INJECTION]: The skill processes external data without safety boundaries. 1. Ingestion points: Markdown research docs and repository source files (SKILL.md). 2. Boundary markers: Absent; there are no instructions to ignore embedded commands in analyzed content. 3. Capability inventory: The skill utilizes
Bash,Write, andEdittools. 4. Sanitization: Absent; content from repositories is processed directly for claim extraction and documentation generation. - [DATA_EXFILTRATION]: The vault detection logic (Phase 0) scans sensitive user directories like
~/Documents,~/notes, and~/vaultsto identify personal knowledge bases (references/vault-integration.md).
Audit Metadata