work-delegator
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from
.pm/backlog/items.yamland.pm/config.yamlto generate instructions for downstream agents (e.g.,builder,tester). - Ingestion points: The
scripts/create_delegation.pyscript reads thetitleanddescriptionof backlog items to build the delegation package. - Boundary markers: The gathered data is placed into a JSON structure, but the skill instructions do not provide explicit delimiters or warnings to the receiving agent to treat the embedded task description as untrusted content.
- Capability inventory: The generated packages are designed for agents with significant capabilities, including file-system modification and command execution.
- Sanitization: No input validation or sanitization is performed on the backlog item content before it is passed to the next agent.
- [COMMAND_EXECUTION]: The skill invokes a local Python script (
scripts/create_delegation.py) using arguments derived from user requests. While the script is part of the skill and usesargparsefor safe argument handling, it constitutes an automated execution surface for local code.
Audit Metadata