apc-orchestrate
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text through the
/apc-orchestrate <goal>command to drive subagent workers. This creates a surface where malicious input could influence the behavior of the orchestrated agents. - Ingestion points: User input provided via the
<goal>parameter in theSKILL.mddescription. - Boundary markers: There are no visible delimiters or 'ignore embedded instructions' warnings for the input text.
- Capability inventory: Orchestrates parallel workers, verifies results, and integrates data, which are powerful capabilities for an agent to exercise based on untrusted input.
- Sanitization: No sanitization or validation of the goal input is present in the skill definition.
- [COMMAND_EXECUTION]: The skill directs the agent to load and follow instructions from a file outside its own directory (
../../.agents/workflows/apc-orchestrate.md). This external dependency dictates the entire operational flow ('worker profiles, wave protocol'), meaning the skill's behavior is entirely dependent on the content of that path.
Audit Metadata