solo-plan

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from user arguments and the local codebase to generate output files. 1. Ingestion points: User-provided task descriptions in $ARGUMENTS and various project files read during the research phase (e.g., CLAUDE.md, docs/prd.md, package.json). 2. Boundary markers: The skill does not utilize explicit delimiters or 'ignore' instructions to isolate ingested content from its internal logic or the output plan templates. 3. Capability inventory: The skill utilizes tools including Bash (for file discovery via find), Write and Edit (for creating plan and spec files), and TaskCreate for pipeline management. 4. Sanitization: There is no evidence of sanitization, escaping, or validation of the ingested external content before it is interpolated into the generated documents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 10:48 AM