coder
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions and a specialized Python evaluation harness (
scripts/run_outcome_eval.py) that executes shell commands via thesubprocessmodule. These commands interact with various AI CLI providers (e.g.,codex,claude,opencode,gemini,pi) and standard tools likegit. - [DYNAMIC_EXECUTION]: The evaluation script (
scripts/run_outcome_eval.py) programmatically generates Python source files (e.g.,slug.py,ports.py,backoff.py) and executes them using theunittestframework. This is used to verify that coding workers have correctly implemented requested features. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external and untrusted data, including repository contents, YAML configuration files (
dna.yaml), and the output of sub-workers. To mitigate this risk, the skill includes explicit 'Iron Law' instructions to treat worker output as unverified claims and uses structured communication protocols (JSON event streams) instead of raw text processing.
Audit Metadata