spec-writer
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides structured guidance for technical specifications, operating solely on local markdown files without network interaction or code execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external requirements and design documents to create implementation-ready specifications. While this creates a surface for indirect instructions, the skill implements sufficient safeguards for its use case.\n
- Ingestion points: Processes user-provided problem statements, PRDs, and existing specifications (e.g.,
prd/SPEC.md).\n - Boundary markers: Implements a
<HARD-GATE>confirmation step and iterative clarification process before document generation to ensure alignment with user intent.\n - Capability inventory: Performs local file reads and writes (specifically to
prd/SPEC.md) based on processed content.\n - Sanitization: Relies on the user review and the structured Symphony-style output format to mitigate unintended instruction extraction from input data.\n- [SAFE]: Recommended patterns in the skill explicitly advocate for secure credential handling by using environment variable placeholders (e.g.,
$VAR_NAME) instead of hardcoded secrets in generated specifications.
Audit Metadata