skills/cloudposse/atmos/atmos-git/Gen Agent Trust Hub

atmos-git

Pass

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Retrieves repository content from remote URIs using the atmos git clone command.
  • [COMMAND_EXECUTION]: Executes Git CLI operations for repository management, including cloning, staging, and committing changes.
  • [PERSISTENCE]: Modifies the repository's .git/hooks directory by installing shims that execute automatically during the Git lifecycle.
  • [DYNAMIC_EXECUTION]: Utilizes a Starlark-based script interpreter to run native commit checks and repository logic defined in external YAML catalogs and local script files.
  • [INDIRECT_PROMPT_INJECTION]: Processes untrusted data from remote repositories and pull requests. The skill implements specific security controls for this surface:
  • Ingestion points: Cloned repository content and Git event payloads (via ATMOS_GIT_HOOK_STDIN).
  • Boundary markers: Implements a 'Fork-PR Trust Gate' and 'Fork Posting Gate' to identify and block unsafe clone and reporting operations in privileged CI environments like pull_request_target.
  • Capability inventory: File system modification (hooks, workdirs), Git networking (push/pull), and Starlark script execution.
  • Sanitization: Employs a temporary Git sandbox for staged checks to prevent modifying the live index during validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 12, 2026, 07:24 AM