planning
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external
ARTIFACT:DISCOVERYsource via theartifact_readtool, creating a potential surface for indirect prompt injection where malicious instructions in the input could influence the planning process. - Ingestion points:
artifact_read(item_id={issue}, artifact_type="feature-context")is used to pull requirements into the agent's context. - Boundary markers: Risk is mitigated by strict structural constraints, including a mandatory markdown output template and behavioral rules such as 'Never invent requirements not present in ARTIFACT:DISCOVERY'.
- Capability inventory: The skill utilizes the
artifact_registerMCP tool to store generated plans. - Sanitization: No explicit sanitization or filtering of the input artifact content is mentioned.
Audit Metadata