skills/polyai/adk/poly-adk-branching/Gen Agent Trust Hub

poly-adk-branching

Pass

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'polyai-adk' package using the 'uv' tool. This is a vendor-owned package necessary for the skill's primary functionality.
  • [DATA_EXFILTRATION]: The skill includes review commands ('poly branch review', 'poly review create') that publish diffs to GitHub Gists. This functionality allows developers to share changes for review using a well-known external service.
  • [CREDENTIALS_UNSAFE]: The documentation notes that a GitHub token must be present in the environment to support the gist-sharing review features. This is a standard requirement for API authentication and does not involve hardcoded secrets.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from remote branches during 'pull' and 'merge' operations.
  • Ingestion points: Data enters the agent's context through remote branch states and incoming files during a three-way merge (SKILL.md).
  • Boundary markers: The tool uses explicit conflict markers (<<<<<<<, =======, >>>>>>>) to delimit local and incoming changes.
  • Capability inventory: The skill executes 'poly' CLI commands for branch management and conflict resolution, including writing to the local filesystem.
  • Sanitization: Files are re-serialized to a canonical form before merging to reduce spurious formatting conflicts, and the merge process follows a structured three-way model.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 15, 2026, 01:06 PM