From maverick
Executes implementation plans step-by-step with verification, failure handling, progress tracking, crash recovery, and acceptance criteria. Adapts behavior for solo vs. guided modes.
How this skill is triggered — by the user, by Claude, or both
Slash command
/maverick:mav-plan-executionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Execute an implementation plan step-by-step. Each step is implemented, verified, and committed before moving to the next. Progress is tracked durably on GitHub so it survives session loss.
Execute an implementation plan step-by-step. Each step is implemented, verified, and committed before moving to the next. Progress is tracked durably on GitHub so it survives session loss.
Two fix attempts, then mav-systematic-debugging, then escalate per mode. This is the single retry budget for implementation work — other skills reference this rule rather than defining their own. If the same fix fails twice, the diagnosis is wrong: step back and think differently, never retry the same change a third time.
Read the task list from one of:
# The tasks comment id lives in the task-progress marker
uv run maverick task-progress read <repo> <issue> # -> .comments.tasks
gh api "repos/<repo>/issues/comments/<id>" --jq '.body'
If resuming after a crash or new session:
- [x] = done, - [ ] = pending)git log — if commits exist for tasks that aren't checked off, the comment update was lost. Check them off now (step 3.6).For each task in the list:
uv run maverick tasks check <repo> <issue> <n>
The command is idempotent, edits exactly one checkbox, and goes
through the App identity when configured. Never PATCH the comment
body by hand.Never batch multiple tasks into one commit unless they are trivially related (e.g. a one-line change and its import).
After all steps are complete, run the project's full verification suite: lint, type checking, all tests. Fix any issues found. Do not proceed to acceptance criteria with failing checks.
Apply the retry budget: diagnose → fix → re-verify, at most twice; then mav-systematic-debugging; then escalate per mode.
| Situation | Solo | Guided |
|---|---|---|
| Step fails after the retry budget | Apply mav-systematic-debugging. If still stuck, ask user for help. | Ask user for help immediately after 2 attempts. |
| Design assumption proves wrong | Reassess against the design. Adjust approach if confident. Only ask user if the change is fundamental. | Pause and discuss with user before adjusting. |
| External blocker (API down, missing dependency) | Document the blocker and ask user. | Document the blocker and ask user. |
| Unsure about implementation approach | Try the most likely approach. If it doesn't work, try the alternative. Ask user only as last resort. | Ask user which approach to take. |
uv run maverick issue comment update --kind plan) and (in guided mode) inform the user.In guided mode, provide brief progress checkpoints at natural break points:
Keep checkpoints brief — one or two sentences. Do not ask for approval to continue unless something went wrong.
After all steps are complete and the full verification suite passes:
Do not proceed to code review until every acceptance criterion is met and all checks pass.
npx claudepluginhub thermiteau/maverick --plugin maverickExecutes written implementation plans: loads/reviews file, runs tasks autonomously (agentic mode) or in batches with pauses (human-in-loop), verifies, reports progress, completes branch.
Executes written implementation plans in small batches with review checkpoints for structured progress tracking. Activate when a plan or spec is ready to implement.
Executes a written implementation plan task-by-task with inline verification, stopping on blockers. Use when a plan exists and you're ready to implement.