delegation-templates
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides instructional templates for delegation and does not contain any malicious code, hidden commands, or unauthorized data access patterns.
- [INDIRECT_PROMPT_INJECTION]: The skill addresses the vulnerability surface where an agent might be influenced by instructions found within the code repositories it processes.
- Ingestion points: Repository content and files specified in parameters like {repo/dir} or {files} are processed by subagents.
- Boundary markers: The templates enforce strict structured output using headers like GOAL, ACCEPTANCE, and REPORT to maintain context separation.
- Capability inventory: The templates are designed for tools that can read/write files and execute search or test commands.
- Sanitization: The 'REVIEW / VERIFY' template explicitly instructs the agent to treat repository-level instructions (e.g., AGENTS.md, CLAUDE.md) as data and forbids them from overriding the review's scope or verdict.
Audit Metadata