workflow-feature-spec

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user requests and external evidence collected from the repository during the investigation phase, creating a potential surface for indirect prompt injection where instructions could be embedded in the data being processed.
  • Ingestion points: External input enters the context through the user request and the data gathered by the $investigate step (SKILL.md).
  • Boundary markers: The workflow implements strict boundary markers through its phase-based structure and explicit quality gates (M1-M7) defined in the 'Required Quality Gates' section (SKILL.md).
  • Capability inventory: The skill has the capability to write to the file system (reports in tmp/reports/ and spec artifacts) and uses spawn_agent for subtasks (SKILL.md).
  • Sanitization: A mandatory review loop ($work-item --mode=review --type=spec-tests --fix-loop) is required to converge on final cases and content, acting as a human-in-the-loop or automated validation filter (SKILL.md).
  • [COMMAND_EXECUTION]: The skill references the execution of a local maintenance script (node .claude/scripts/lib/workflow-skill-contract.cjs) to regenerate workflow blocks, which is a standard development task for this workflow framework (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 08:16 PM