spec
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses Git commands like 'git rev-parse' and 'git diff' to establish codebase baselines and detect drift throughout the specification process.
- [EXTERNAL_DOWNLOADS]: The skill facilitates the cloning of external Open Source Software (OSS) repositories to a local directory for deep technical analysis and research purposes.
- [PROMPT_INJECTION]: The skill processes untrusted data from web searches and external repositories, creating an attack surface for indirect prompt injection.
- Ingestion points: Web search results and external OSS repository source code.
- Boundary markers: The skill enforces an 'evidence-driven' stance and requires validating assumptions against primary sources.
- Capability inventory: The skill has access to shell commands (Git, Bash), file system writes (SPEC.md and evidence files), and the ability to spawn nested agent instances via the '/nest-claude' tool.
- Sanitization: No explicit sanitization or filtering of external research content is documented in the instructions.
Audit Metadata