improve-codebase-architecture-project
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
gh issue listcommand to retrieve a list of existing architectural proposals from GitHub to ensure novelty in its own recommendations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from existing GitHub issues and codebase files. This information is processed to generate new architectural proposals without explicit sanitization or boundary markers to ignore embedded instructions in the source content. The resulting output is used by a downstream driver script to create new GitHub issues.
- Ingestion points:
gh issue listoutput and repository files accessed via theExploretool. - Boundary markers: None present.
- Capability inventory: The skill generates a structured JSON output containing a PRD title and body, which is intended for a driver script to execute
gh issue create. - Sanitization: No sanitization of ingested content before interpolation into the PRD output.
Audit Metadata