The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the omp binary and the git CLI using the scripts/relay.mjs script. For write-capable tasks, the relay script passes the --yolo flag to the omp binary, which automatically approves all tool calls (including shell commands and file system modifications), bypassing manual user review at the tool level.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a relay for external task briefs that can influence the behavior of the implementer agent.
Ingestion points: Task instructions are ingested through stdin or the --brief argument in scripts/relay.mjs.
Boundary markers: The relay script does not programmatically wrap the brief in delimiters or isolation markers; it relies on the user to follow formatting advice in the documentation.
Capability inventory: The target omp tool is a high-capability agent with the ability to read and write files and execute arbitrary shell commands.
Sanitization: No sanitization or filtering is performed on the brief content before it is passed to the agent's input stream.
[DYNAMIC_EXECUTION]: The scripts/relay.mjs helper dynamically executes an external script (lane.mjs) from a relative path (../../delegate-setup/scripts/lane.mjs) outside the skill folder. It also uses the node:child_process module to launch external binaries with arguments derived from user input, although it implements basic token validation for certain flags.

omp-delegate