maotoumao/Cebian / .agents/skills/start-task
start-task
VerifiedResume execution of an approved plan from the next unchecked subtask, following the gated Task Execution Workflow in AGENTS.md.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.3 KB |
| agents/openai.yaml | config | 229 B |
SKILL.md
View on GitHub---
name: start-task
description: "Resume execution of an approved plan from the next unchecked subtask, following the gated Task Execution Workflow in AGENTS.md."
---
Resume work on an in-progress plan. Follow the **Task Execution Workflow** defined in [AGENTS.md](../../../AGENTS.md) strictly — one subtask, code review, test instructions, then stop.
## Steps
1. **Locate the plan**
- If the user provided a path, use that file.
- Otherwise list [docs/plans/](../../../docs/plans/) and pick the most recently modified `*.md` plan that is not marked `PROPOSAL`. Confirm the choice with the user in one short line before proceeding.
2. **Find the next subtask**
- Read the plan and locate the first `## Task N:` section that still contains unchecked `- [ ]` steps.
- If every task is fully checked, report "All tasks complete" and stop. Do not invent new work.
3. **Sync the TODO list**
- Use your harness's todo/task list to mirror the plan's tasks. Mark already-completed tasks as `completed`, the located next task as `in-progress`, and remaining tasks as `not-started`.
4. **Execute exactly that one subtask**
- Implement only the steps inside that single `## Task N:` section.
- Do not touch files belonging to later tasks. Do not look ahead.
- When the implementation is done, update the plan file to check off (`- [x]`) the steps you completed.
5. **Code review**
- Run the repository's `code-review` workflow in the current harness's dedicated read-only reviewer. Pass it the files changed in this task and the plan section as context, then address any actionable issues before continuing.
6. **Hand off to the user**
- Provide concrete manual testing instructions: what to run (`pnpm dev` is usually already running in the `dev` terminal), what to open, what to click, expected console output, expected UI state.
- Then **stop**. Do not start the next task. Do not mark the TODO `completed` yet.
7. **Wait for explicit approval**
- Only after the user replies with explicit approval (e.g., "approved", "next", "looks good", "passed") should the TODO be marked `completed`. Acknowledgements and silence are not approval.
If the plan turns out to be wrong or incomplete during step 4, stop and trigger the **Plan-First** re-approval flow instead of pushing forward.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |