coordinator
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command (
ls -1 "$POWERSUIT_OVERLAY/ai-ops") to check for the existence of local configuration overlays using an environment variable. - [INDIRECT_PROMPT_INJECTION]: The coordinator workflow involves ingesting and acting upon artifacts and reports produced by separate 'worker' sessions, which could contain instructions designed to manipulate the agent's behavior.
- Ingestion points: Worker reports, file diffs, task ledger entries (
work/ledger.md), and generated artifacts. - Boundary markers: The instructions require the use of 'briefs' that name specific goals and restricted file paths for workers.
- Capability inventory: File system writing, management of worker agent sessions, and integration actions such as committing, merging, or publishing code.
- Sanitization: The skill includes explicit rules for artifact verification, requiring the agent to review diffs and run test outputs rather than relying solely on worker reports.
Audit Metadata