dev-improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Apr 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection attacks where instructions embedded in project documentation could influence the agent's behavior.
- Ingestion points: The skill reads external codebase documentation, specifically
CONTEXT.md,CONTEXT-MAP.md, and Architecture Decision Records (ADRs) within thedocs/adr/directory to extract domain vocabulary and architectural context (SKILL.md, Step 1). - Boundary markers: There are no explicit instructions or delimiters used when reading these files to differentiate between data and potentially malicious instructions.
- Capability inventory: The skill utilizes the
Exploreagent tool to crawl the codebase and thegh(GitHub CLI) to create issues. It also performs agit remotecheck (SKILL.md, Step 4). - Sanitization: There is no evidence of sanitization or validation of the domain vocabulary or context extracted from project files before it is processed or included in the final GitHub issue bodies.
Audit Metadata