writing-plans
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data (project specifications and design documents) to generate implementation plans. This creates a potential surface for indirect prompt injection if the specifications contain malicious instructions.
- Ingestion points: The skill reads external files via the
Spec: [path]field in the plan header and during the plan generation process. - Boundary markers: The skill does not currently mandate explicit delimiters (like XML tags or specific 'ignore instructions' warnings) when referencing the source spec in the generated output.
- Capability inventory: The resulting plans generated by this skill include file creation/modification paths, shell commands for testing (
pytest), and git operations (git commit). - Sanitization: No explicit sanitization or validation of the specification content is mentioned in the instructions.
Audit Metadata