req-ship

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external sources such as Lark documents and Meegle work items, which creates a surface for indirect instructions.
  • Ingestion points: The lark-cli docs +fetch command in SKILL.md retrieves content from external document URLs.
  • Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard instructions embedded within the fetched document content.
  • Capability inventory: The agent has the ability to execute shell commands (meegle workflow transition-state) to change the status of work items.
  • Sanitization: No explicit sanitization or validation of the fetched document content is mentioned before it is used to inform the checklist.
  • Mitigation: The skill mandates a human-in-the-loop step: 'Before transitioning, show the command and target state to the requester for confirmation.'
  • [COMMAND_EXECUTION]: The skill relies on shell command execution to interact with internal project management tools.
  • Evidence: Uses meegle and lark-cli for project search, metadata retrieval, and workflow transitions.
  • Risk: These commands interpolate variables such as project keys, work item IDs, and document URLs, which could be exploited if inputs are not properly handled by the underlying execution environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 02:36 AM