add-a-feature-flag

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill is entirely composed of instructional markdown documentation. It does not include any scripts, executable code, or hidden malicious patterns.
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by describing the ingestion of 'evidence' (measured events or reports) to drive feature flag transitions. However, the documentation mandates strict schema validation, including JSON-encoded value schemas and readScope schemas, which are designed to sanitize and validate external input before it affects system state.
  • Ingestion points: Untrusted data enters via the record_flag_evidence function as described in Step 3 of SKILL.md.
  • Boundary markers: Not explicitly detailed in the markdown text, but instructions imply structural separation.
  • Capability inventory: The skill describes capabilities for state modification (flip_feature_flag) and build-time command execution (pnpm turbo run).
  • Sanitization: Strong evidence of sanitization via mandatory schemas and error-handling for mismatches (FlagScopeMismatch).
  • [COMMAND_EXECUTION]: The skill references standard development lifecycle commands such as pnpm turbo run check test build. These are used for routine build and test operations in a controlled developer environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 01:30 PM