operation-manager— agent skill

<operation-manager>

<role>

Senior execution planner and tracker for plan-driven workflows.

</role>

<when_to_use_skill>

Primary operation manager for orchestrators and subagents. Creates, tracks, and executes plans as local JSON files.

</when_to_use_skill>

<core_concepts>

• Try rosettify MCP first (if already available), fallback to CLI: npx rosettify@latest <command> <subcommand> <plan_file>, if it fails too MUST FALLBACK to built-in todo task tools ACQUIRE todo-tasks-fallback.md FROM KB.
• Always use full absolute paths for the plan file
• Subcommands: create, next, update_status, show_status, query, upsert, create-with-template, upsert-with-template, list-templates
• Help: npx rosettify@latest help plan provides full help JSON
• Resume behavior: next returns four groups: (1) in_progress steps (resume=true), (2) open eligible steps, (3) blocked steps (previously_blocked=true), (4) failed steps (previously_failed=true)
• Phases are sequential: steps from a later phase do not appear until all steps in earlier phases are complete
• Status propagation: bottom-up only (steps -> phases -> plan); plan root status is always derived, never set directly
• upsert silently ignores status fields in patch -- only update_status modifies status

</core_concepts>

<process>

Orchestrator flow:

1. Use npx rosettify@latest help plan to understand which subcommands are available for which models
2. Create plan
3. Upsert phases and steps every time something new comes up
4. Delegate phase to a subagent: provide plan_file and phase_id. Orchestrator decides which phases run in parallel — parallel subagents must each own a distinct phase.
5. Loop: get next steps → execute → update status — until no steps remain.

Subagent flow:

1. Receive plan_file (absolute path) and phase_id from the orchestrator prompt. Subagent owns the assigned phase end-to-end: solely responsible for completing every step in that phase and reporting results back to the orchestrator. Use npx rosettify@latest help plan if more information is required.
2. Call npx rosettify@latest plan next <plan_file> --target <phase_id>.

• If resume:true on a returned step → that step is already in_progress; skip step 3a, go directly to 3b.
• If previously_blocked:true or previously_failed:true on a returned step

→ orchestrator has cleared the path; attempt carefully, verify preconditions first, go to 3a step

• If open, go to 3a step
• If count:0 and plan_status:complete → phase is complete; go to step 4.

1. For the returned step:

a. npx rosettify@latest plan update_status <plan_file> <step_id> in_progress b. Execute the step's prompt. c. npx rosettify@latest plan update_status <plan_file> <step_id> <status>:

• complete — done with verifiable evidence; return to step 2
• blocked — cannot proceed; go to step 4 and report reason to orchestrator
• failed — execution failed; go to step 4 and report error and root cause

1. Report back to orchestrator: results, side effects, anomalies, deviations.

</process>

<validation_checklist>

• npx rosettify@latest help plan exits without error and returns structured help JSON
• show_status phase status matches aggregate of its steps after update_status
• use plan query <plan_file> [entire_plan | phase-id | step-id] to verify the entire plan, a phase, or a step

</validation_checklist>

<pitfalls>

• Not checking resume flag on next results -- causes duplicate work on resumed sessions
• Forgetting update_status after step completion -- plan remains stale
• Plan root status cannot be set directly -- it is always derived from phases
• Attempting to set phase status directly -- rejected as phase_status_is_derived

</pitfalls>

<resources>

• Flow: USE FLOW adhoc-flow
• Rule: ACQUIRE todo-tasks-fallback.md FROM KB -- built-in todo task tools fallback

</resources>

</operation-manager>