to-spec
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill instructions do not contain any patterns typical of prompt injection or jailbreak attempts. Directives like 'freeze' and 'stop' are used to control the agent's logical flow for technical specification writing rather than to override safety filters.
- [DATA_EXFILTRATION]: No network operations or sensitive data access patterns were found. The skill operates on local repository artifacts and generates markdown files. Test fixtures included in the skill use dummy values for credentials (e.g., 'pub_example').
- [REMOTE_CODE_EXECUTION]: The skill is designed to produce documentation (implementation specs) and does not involve the execution of shell commands, scripts, or external binaries.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted data such as task descriptions and code, it includes defensive procedures like the 'UNVERIFIED' label for evidence gaps and a mandatory validation step. Human review is explicitly required before any implementation or merging occurs, providing a strong safety boundary.
- [DYNAMIC_CONTEXT_INJECTION]: The skill does not use the dynamic context injection syntax (`!``command```).
Audit Metadata