plan-issues
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and assess external codebase content and documentation, creating a surface where malicious instructions embedded in a scanned repository could attempt to influence the agent's output.
- Ingestion points: External repository content, configuration files, and policies as defined in SKILL.md and compliance-review/REFERENCE.md.
- Boundary markers: While the instructions emphasize explicit criteria and acceptance rules, they do not prescribe technical delimiters (e.g., XML tags or specific markers) to separate untrusted codebase data from the agent's internal logic.
- Capability inventory: The skill includes the capability to 'create only validated selected drafts' in a hosted issue backlog once authorization is granted (SKILL.md).
- Sanitization: The skill implements a comprehensive quality validation protocol (references/vstack/protocols/quality-validation.md) requiring self-assessment, independent review, and verification against proof, which acts as a primary mitigation against deceptive or erroneous input.
Audit Metadata