The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core functionality relies on spawning the qodercli and git binaries via Node.js child_process.spawn and spawnSync to execute coding tasks and inspect the repository state.
[INDIRECT_PROMPT_INJECTION]: As an orchestration skill, it processes 'briefs' (instructions) and workspace context which could serve as vectors for indirect prompt injection if they contain untrusted data.
Ingestion points: Task briefs provided to the relay script and files within the project workspace.
Boundary markers: The documentation recommends using XML-like tags (e.g., <task>, <action_safety>) to structure and delimit instructions within the brief.
Capability inventory: The skill has the capability to modify the local filesystem and run project-specific commands through the delegated CLI agent.
Sanitization: The skill mandates a 'Review and Land' workflow where the orchestrator must independently verify all diffs, re-run project gates, and manually commit changes, preventing the delegated agent from directly affecting the repository's history.
[DYNAMIC_EXECUTION]: The skill utilizes a bundled Node.js script (scripts/relay.mjs) to orchestrate the lifecycle of the external coding agent. The script includes logic for handling process timeouts, signal termination, and structured log parsing.

qoder-delegate