Delegates tasks to OMP via the omp CLI in non-interactive JSON mode.
npx skills add https://github.com/akhy/agent-skills --skill omp-delegate-akhy
Please help me install this Agent Skill. Skill: omp-delegate Source: https://github.com/akhy/agent-skills/tree/main/omp-delegate Command: npx skills add https://github.com/akhy/agent-skills --skill omp-delegate-akhy
omp CLI and capture structured results without a TUI. ## Core Features & Use Cases - Fire-and-forget delegation: Run a prompt with --no-session --mode json for ephemeral, stateless execution. - Blocking result capture: Pipe JSON output through jq to extract the final assistant message from the agent_end event. - Multi-shot sessions: Create a session with --session-dir, capture the session ID, and resume it with --resume for retained context across calls. - Use Case: An orchestrating agent needs a code change reviewed by OMP. It runs omp -p "review this diff" --no-session --mode json --auto-approve, extracts the final text with jq, and continues its own workflow with the result. ## Quick Start Ask the agent to delegate a prompt to OMP by running omp with the -p flag in JSON mode and returning the final assistant message.High-intent search queries and answers about installing and using this skill.
Run omp with the -p flag for the prompt, plus --no-session and --mode json for ephemeral structured output. Add --auto-approve to skip tool-approval prompts so the command completes without terminal interaction.
Pipe the omp output through jq filtering for the agent_end event, then select assistant messages with text content and take the last entry. This strips noisy intermediate events and returns only the final response text.
Yes. Create the first call with --session-dir pointing to a temp directory, parse the session ID from the first JSON line, then pass --resume with that ID on subsequent calls. Using --continue would create a new session instead.
Exit code 0 means success, 1 is a general error, 2 indicates a usage error, and 126 means the omp binary was not found. Check these codes after each delegation to detect failures programmatically.
No. The --thinking off flag only suppresses TUI display; thinking tokens are still generated by the model. Stripping thinking from the output stream saves caller-side tokens, not model-side tokens.
brikkoAIExecute one-off Gemini CLI prompts with optional model and output configuration.
mrsono0Analyze Git status and diffs to generate session wrap-up reports.
obraCoordinate multiple Claude Code sessions in tmux for parallel task delegation.
archibateAutomate headless Claude Code CLI, stream-json, and resume modes.