factory-deploy
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external trackers and issue descriptions which could contain malicious instructions.\n
- Ingestion points: External data is ingested via the
mr_urlinput and read from 'tracker bodies' or 'linked MR/issue text' in workflow steps such asverify,gate-merge,gate-canary-expand, andgate-full-expand.\n - Boundary markers: The skill explicitly instructs the agent to treat these sources as data rather than instructions, using the 'message-is-data' invariant (Invariant 6) and specific ignore-directive prompts.\n
- Capability inventory: The skill has high-privilege capabilities including merging code to repositories, creating SemVer tags, and managing production rollout percentages.\n
- Sanitization: No programmatic sanitization or escaping is performed on the ingested text; the skill relies on the AI following instruction-based boundaries to mitigate injection risks.
Audit Metadata