orchestrate
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes local shell and PowerShell scripts (e.g.,
scripts/sync-agents.sh,scripts/build-and-install.ps1,scripts/codex/apc-codex-run.sh) to manage the build environment and subagent synchronization. This behavior is consistent with the skill's role as a manager of development workflows within an APC checkout.\n- [INDIRECT_PROMPT_INJECTION]: The skill constructs subagent briefs by interpolating user-provided goals and project metadata from files likestatus.json.\n - Ingestion points: User-provided goal strings and plugin-specific status files.\n
- Boundary markers: The skill references a 'Worker brief contract' and explicit file boundaries to limit subagent scope, although specific escaping mechanisms for untrusted content are not detailed.\n
- Capability inventory: The skill is capable of spawning write-capable subagents and executing local administrative and build scripts.\n
- Sanitization: No specific sanitization or filtering of input data is mentioned in the instructions.\n- [DYNAMIC_EXECUTION]: The skill generates per-host adapters and routes tasks to different worker tiers at runtime using local scripts, facilitating dynamic coordination of specialized agents.
Audit Metadata