spec-driven-apply

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local utility script spec-driven.js via Node.js to modify and apply change sets.
  • [COMMAND_EXECUTION]: The agent is instructed to run arbitrary project commands such as linting and unit tests to verify implementation tasks.
  • [PROMPT_INJECTION]: The skill contains highly emphatic instructions directing the agent to ignore all prior conversational context. This is implemented as a reliability measure to ensure the agent uses the repository files as the sole source of truth for the change specification.
  • [PROMPT_INJECTION]: Indirect prompt injection surface exists as the skill reads and acts upon tasks and instructions defined in repository files.
  • Ingestion points: Files within the .spec-driven/ directory (e.g., tasks.md, proposal.md).
  • Boundary markers: Absent; the agent is instructed to treat file content as authoritative instructions.
  • Capability inventory: File system modification, local script execution, and project test execution.
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 01:18 PM