specification-writing
Pass
Audited by Gen Agent Trust Hub on Apr 19, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill demonstrates a surface for indirect prompt injection as it processes untrusted local data and incorporates it into workflows involving external model calls.
- Ingestion points: Content is read from 'patent/CLAIMS.md', 'patent/INVENTION_DISCLOSURE.md', and 'patent/PRIOR_ART_REPORT.md'.
- Boundary markers: The skill does not implement delimiters or 'ignore embedded instructions' warnings when passing processed data to the 'mcp__codex__codex' tool or sub-skills.
- Capability inventory: The agent has access to 'Bash', 'Write', 'Agent', and 'mcp__codex__codex' (external model interface).
- Sanitization: There is no evidence of validation or filtering for input content before it is processed.
- [COMMAND_EXECUTION]: The skill utilizes the 'Bash' tool to execute shell commands for creating the patent directory structure and writing specification sections via heredocs.
Audit Metadata