spec-first-planning

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes evaluation and testing scripts (eval/run_eval.py, assets/test_spec_lint.py, assets/test_spec_to_tasks.py) that utilize the subprocess module to execute the bundled Python linter and compiler. These executions are performed in a controlled manner against test fixtures and temporary files for verification purposes.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a processing pipeline for user-supplied markdown specifications, which constitutes an ingestion surface for untrusted data.
  • Ingestion points: Markdown specification files are ingested by assets/spec_lint.py and assets/spec_to_tasks.py via command-line arguments.
  • Boundary markers: The tools rely on specific structural markers, such as H2 headers (e.g., ## Requirements) and strict ID prefixes (e.g., - R1:), to identify valid input.
  • Capability inventory: The skill's scripts are deterministic text parsers that do not perform network operations, sensitive file access, or persistent system modifications.
  • Sanitization: Input is validated against a strict set of grammar rules (modal obligations, metric requirements, and ID sequencing) using regular expressions, effectively filtering out non-conforming content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 07:06 PM
Security Audit — agent-trust-hub — spec-first-planning