manage-task-checkpoints

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Uses the cosh-gateway CLI tool to perform task and snapshot operations. This tool appears to be a specialized component of the platform environment as specified in SKILL.md.- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external task data and snapshot states, which constitutes an ingestion surface for potentially untrusted content.
  • Ingestion points: Results from cosh-gateway task snapshot list, preview, and diff commands described in SKILL.md.
  • Boundary markers: The instructions mandate natural-language confirmation from the user and clear descriptions of file effects before any workspace switch occurs.
  • Capability inventory: The skill is capable of executing specific shell commands via the cosh-gateway tool for snapshot management.
  • Sanitization: The skill explicitly instructs the agent to accept only complete IDs returned by the gateway and prohibits guessing or prefix expansion.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 03:58 AM