bugfix-delivery
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved from the Feishu Project management system.
- Ingestion points: Data is ingested via the
meegle workitem getcommand in Step 1, which retrieves bug titles, descriptions, and reporter names. - Boundary markers: The instructions lack explicit boundary markers or "ignore instructions" directives when interpolating this bug data into prompts for branch naming, MR descriptions, or bug comments.
- Capability inventory: The skill has the authority to execute
gitcommands, create Merge Requests, modify local worktrees viawt, and post automated comments to the project management system. - Sanitization: There is no mention of sanitizing or escaping the retrieved bug content before it is used to generate command-line arguments or downstream sub-agent instructions.
Audit Metadata