dev-plan
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from user-provided PRDs and development plans.\n
- Ingestion points: User-supplied requirements, PRD documents, and existing development plans provided via the chat context or referenced files (e.g., SKILL.md, plan-skeleton.md).\n
- Boundary markers: The instructions mandate a "System Architect" role that treats plan contents as claims to be verified against code facts. It uses specific placeholders and structural requirements defined in
references/plan-skeleton.mdto delimit content and instructs implementation agents to ignore metadata during execution.\n - Capability inventory: The skill possesses capabilities to read local files, execute provided validation scripts (scripts/check_paths.sh, scripts/check_plan.py, scripts/check_progress.sh), and inspect Git repository status (
git status).\n - Sanitization: The skill employs automated validation scripts that verify the existence of referenced paths and check the internal consistency of generated plans. The shell scripts include regex filters (e.g., in
check_paths.sh) to prevent command injection during path processing.
Audit Metadata