complete-task

name	complete-task
description	Mark a task as completed and move it to done. Use when the user finishes work, says "done with task", "complete task", "mark as done", or when all acceptance criteria are met.
argument-hint	["task ID or number"]
allowed-tools	["Read","Write","Bash","Glob","Grep"]
model	haiku

Complete Task

Mark a task as completed, move it to the complete directory, and handle learnings.

Arguments

Task identifier (optional): $ARGUMENTS — accepts TSK-001, 001, or 1.

Prerequisites

Verify .backlog/ exists.

Workflow

1. Identify Task

If $ARGUMENTS provided, find matching file in .backlog/tasks/active/. If not there, check .backlog/tasks/pending/ (a task can be completed without having been started) — if found there, confirm with the user before proceeding. If found nowhere, say so and stop.

If no argument, list active tasks and ask which to complete.

2. Verify Acceptance Criteria

Read the task file and check acceptance criteria:

• If all are checked [x], proceed
• If unchecked criteria exist, warn the user:

TSK-{{ID}} has unchecked acceptance criteria:
- [ ] Criterion 3

Options:
1. Mark as complete anyway (criteria no longer relevant)
2. Continue working (abort completion)
3. Update criteria (remove/modify items)

3. Extract Learnings

Gather learning candidates from two sources, then create each one silently using auto-draft mode (see /flowstate:add-learning Step 2a). Do NOT ask the user first.

3a. From the task file's Learnings section — each entry becomes a full learning (expand the one-line entry into Context / Insight / Application using the task body and tags).

3b. From the recent conversation — scan for non-obvious discoveries that emerged while working on this task: gotchas, root causes ("ah, that failed because…"), patterns that worked, things to avoid. High signal only — skip routine work.

For every candidate from 3a or 3b, run:

cat <<'BODY' | node "${CLAUDE_PLUGIN_ROOT}/dist/bin/flowstate.js" learning-create --title "{{TITLE}}" --tags "{{TAGS}}" --task {{TSK_ID}} --body -
{{BODY_DRAFTED_FROM_CONTEXT}}
BODY

Surface every captured learning in the confirm output (Step 5) so the user can edit or delete if signal was wrong. If no candidates from either source, skip silently — do NOT prompt.

4. Complete Task via CLI

node "${CLAUDE_PLUGIN_ROOT}/dist/bin/flowstate.js" task-move {{ID}} --to complete

The CLI updates frontmatter (status: complete, completed: today), adds a progress log entry, moves the file to tasks/complete/, and updates tasks/index.md automatically.

5. Confirm Completion

Completed TSK-{{ID}}: {{TITLE}}
{{PENDING_COUNT}} tasks remaining.

Learnings auto-captured: {{N}}
- LRN-XXX: {{TITLE}}            ← list each, so the user can edit/delete

/flowstate:next-task — Get a recommendation · /flowstate:add-learning — Capture another

If N == 0, omit the "Learnings auto-captured" block entirely.