writing-plans
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingest external specifications or requirements to produce implementation plans that contain shell commands and code snippets. These plans are explicitly designed to be executed by other agents (e.g., using
superpowers:subagent-driven-development), creating a multi-step execution chain where malicious instructions in the input spec could influence subsequent system actions. - Ingestion points: External project specifications or requirements documents mentioned in the description and header templates.
- Boundary markers: The skill lacks explicit delimiters or instructions to treat the spec content as untrusted data during processing.
- Capability inventory: The generated output (saved to
docs/superpowers/plans/) includes shell commands for testing (pytest) and version control (git), as well as functional code blocks for implementation. - Sanitization: The instructions mandate copying values verbatim from the spec (e.g., Global Constraints) without validation or sanitization, allowing potential instruction injection into the plan header.
Audit Metadata