From ai-dev-assistant
Finalizes tasks by moving the task folder to completed/, updating project_state.md, and running 5 quality gates that block completion on failure.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ai-dev-assistant:task-completerinheritThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Finalize tasks and update project memory.
Finalize tasks and update project memory.
Load before completing any task:
| Reference | Enforces |
|---|---|
references/quality-gates.md | Gates 2, 3, 4 (must ALL pass) |
framework security guide (use /dev-guides-navigator security to resolve for the project's stack) | Gate 4 security review |
Activate when you detect:
/ai-dev-assistant:complete commandTask CANNOT be completed until ALL gates pass:
| Gate | Check | Blocking? |
|---|---|---|
| Gate 1 | Code standards (invoke code-pattern-checker) | YES |
| Gate 2 | Tests pass (user confirms) | YES |
| Gate 3 | Architecture compliance | YES |
| Gate 4 | Security review (invoke guide-integrator with the framework's security topic, resolved via /dev-guides-navigator security; do NOT WebFetch directly) | YES |
| Gate 5 | All acceptance criteria in task.md are [x] and ## Phase Status shows Phases 1–3 complete | YES |
Use Read on the task tracker file: {project_path}/implementation_process/in_progress/{task_name}/task.md (v3.0.0 folder structure; task.md lives inside the task folder).
Check each acceptance criterion. Ask user:
Completion checklist for {task_name}:
Acceptance Criteria:
- [ ] {criterion 1} - Is this done?
- [ ] {criterion 2} - Is this done?
- [ ] {criterion 3} - Is this done?
Confirm all acceptance criteria are met (yes/no):
If NO, identify what's remaining and continue working.
ALL gates must pass before completion:
Invoke code-pattern-checker skill on modified files.
Ask user to confirm:
Tests verification (user must run):
<project test command> {test_path}
- [ ] All existing tests pass?
- [ ] New code has test coverage?
- [ ] No skipped tests without documented reason?
Confirm tests pass (yes/no):
Check against architecture/main.md:
guide-integrator with the framework's security topic (resolved via /dev-guides-navigator security; do NOT WebFetch directly)task.md are [x]## Phase Status in task.md shows Phases 1, 2, 3 all [x]research.md, architecture.md, implementation.md all present in the task folderIf ANY blocking gate fails: Task completion is BLOCKED. Fix issues first.
Use Edit to add completion section to the task file:
---
## Completion
**Completed:** {YYYY-MM-DD}
**Final Status:** Complete
### Summary
{Brief description of what was implemented}
### Files Changed
| File | Action |
|------|--------|
| src/... | Created |
| tests/... | Created |
| config/... | Modified |
### Test Results
- Unit tests: {count} passing
- Integration tests: {count} passing
- Total: All passing
### Notes
{Any implementation notes, deviations, or decisions made}
v3.0.0 tasks are folders (task.md + research.md + architecture.md + implementation.md + any component files). Move the entire folder, not a single file.
mkdir -p "{project_path}/implementation_process/completed"
mv "{project_path}/implementation_process/in_progress/{task_name}" "{project_path}/implementation_process/completed/{task_name}"
({task_name} is the folder name — no .md suffix. The folder contains task.md and all phase artifacts.)
Use Edit to update:
## Progress
### Completed Tasks
| Task | Completed | Notes |
|------|-----------|-------|
| {task_name} | {date} | {one-line summary} |
## Current Focus
{Update to next task or "Ready for next component"}
Use Bash to list remaining in-progress task directories (v3.0.0 folder structure — tasks are directories, not .md files):
ls -1d "{project_path}/implementation_process/in_progress/"*/ 2>/dev/null
Each result is a task folder; read its task.md to inspect ## Phase Status and acceptance criteria before ranking.
Analyze dependencies and priorities. Present:
Task complete: {task_name}
Next task options:
1. {next_task} - {reason: dependency unblocked / priority}
2. {alternative} - {reason}
3. No more tasks - component complete
Which task next? (1/2/3 or other):
If this was the last task for a component, suggest:
Component {name} appears complete.
Run final validation?
- superpowers:verification-before-completion
- Full test suite
- Integration tests
Proceed? (yes/no)
STOP and wait for user:
npx claudepluginhub camoa/claude-skills --plugin ai-dev-assistantMarks tasks as completed after verifying subtasks, acceptance criteria, optional verify checks (bash/assert), and worklogs. Handles solo and PR-review workflows.
Close one completed or parked task. Use when the user asks to finish, mark done, defer, park, drop, or archive a task. Set finished or deferred, move the file to archive, update links, and relint.
Executes all remaining tasks in sequence until completion, stopping on first failure. Includes pre-flight recommendation for Sonnet/Opus at high effort and monorepo project context resolution.