Implementing a Task

Execute the implementation plan, track progress, and complete the task.

Announce at start: "I'm using the knowns.task.implement skill to implement task [ID]."

Core principle: CHECK AC ONLY AFTER WORK IS DONE.

The Process

Step 1: Review Current State

{{#if mcp}}

mcp__knowns__get_task({ "taskId": "$ARGUMENTS" })

{{else}}

knowns task $ARGUMENTS --plain

{{/if}}

Verify:

• •Plan exists and is approved
• •Timer is running
• •Know which ACs are pending

Step 2: Check for Applicable Templates

Before writing code, check if there's a template that matches:

knowns template list

If template exists:

1. •Read linked doc for context
2. •Use template to generate boilerplate
3. •Customize generated code as needed

{{#if mcp}}

// Read template's linked doc
mcp__knowns__get_doc({ "path": "<template-doc>", "smart": true })

# Generate code from template (reduces context, ensures consistency)
knowns template run <template-name> --name "MyComponent"

{{else}}

# Read template's linked doc
knowns doc "<template-doc>" --plain

# Generate code from template (reduces context, ensures consistency)
knowns template run <template-name> --name "MyComponent"

{{/if}}

Why use templates:

• •Reduces context (no need to generate boilerplate)
• •Ensures consistency with project patterns
• •Faster implementation

Step 3: Work Through Plan

For each step in the plan:

1. •Check for template (use if available)
2. •Do the work (generate or write code)
3. •Check related AC (only after work is done!)
4. •Append progress note

{{#if mcp}}

// After completing work for AC #1:
mcp__knowns__update_task({
  "taskId": "$ARGUMENTS",
  "checkAc": [1],
  "appendNotes": "✓ Done: brief description"
})

{{else}}

# After completing work for AC #1:
knowns task edit $ARGUMENTS --check-ac 1
knowns task edit $ARGUMENTS --append-notes "✓ Done: brief description"

{{/if}}

Step 4: Handle Scope Changes

If new requirements emerge during implementation:

Small change: {{#if mcp}}

mcp__knowns__update_task({
  "taskId": "$ARGUMENTS",
  "addAc": ["New requirement"],
  "appendNotes": "⚠️ Scope: added requirement per user"
})

{{else}}

knowns task edit $ARGUMENTS --ac "New requirement"
knowns task edit $ARGUMENTS --append-notes "⚠️ Scope: added requirement per user"

{{/if}}

Large change:

• •Stop and ask user
• •Consider creating follow-up task
• •Update plan if needed

Step 5: Verify & Complete

When all ACs are checked:

1. Verify code quality:

npm test        # or project's test command
npm run lint    # or project's lint command
npm run build   # if applicable

Don't complete if verification fails. Fix issues first.

2. Add implementation notes (REQUIRED for audit):

Document all changes made for audit trail. Use appendNotes to preserve history:

{{#if mcp}}

mcp__knowns__update_task({
  "taskId": "$ARGUMENTS",
  "appendNotes": "## Implementation Complete\n\n### Files Changed\n- `src/path/file.ts` - Added X\n- `src/path/other.ts` - Modified Y\n\n### Key Changes\n- Change 1: description\n\n### Testing\n- Test coverage / manual testing done"
})

{{else}}

knowns task edit $ARGUMENTS --append-notes $'
## Implementation Complete

### Files Changed
- `src/path/file.ts` - Added X
- `src/path/other.ts` - Modified Y
- `tests/file.test.ts` - Added tests

### Key Changes
- Change 1: description
- Change 2: description

### Testing
- Test coverage / manual testing done
'

{{/if}}

IMPORTANT: Always use appendNotes (not notes) to preserve audit trail.

3. Stop timer and mark done:

{{#if mcp}}

mcp__knowns__stop_time({ "taskId": "$ARGUMENTS" })

mcp__knowns__update_task({
  "taskId": "$ARGUMENTS",
  "status": "done"
})

{{else}}

knowns time stop
knowns task edit $ARGUMENTS -s done

{{/if}}

Step 6: Consider Knowledge Extraction

If generalizable patterns were discovered:

/knowns.extract $ARGUMENTS

Progress Tracking

Use concise notes:

{{#if mcp}}

// Good
mcp__knowns__update_task({
  "taskId": "$ARGUMENTS",
  "appendNotes": "✓ Auth middleware implemented"
})

// Bad (too verbose)
mcp__knowns__update_task({
  "taskId": "$ARGUMENTS",
  "appendNotes": "I have successfully completed..."
})

{{else}}

# Good
knowns task edit $ARGUMENTS --append-notes "✓ Auth middleware implemented"

# Bad (too verbose)
knowns task edit $ARGUMENTS --append-notes "I have successfully completed..."

{{/if}}

Completion Checklist

• • All ACs checked
• • Tests pass
• • Lint clean
• • Implementation notes added (with file changes for audit)
• • Timer stopped
• • Status set to done
• • Knowledge extracted (if applicable)

Red Flags

You're doing it wrong if:

• •Checking AC before work is actually complete
• •Making changes not in the approved plan (without asking)
• •Skipping tests
• •Not tracking progress with notes
• •Marking done without verification

When to Stop

STOP and ask when:

• •Requirements unclear or contradictory
• •Approach isn't working after 2-3 attempts
• •Need changes outside approved scope
• •Hit unexpected blocker

If Verification Fails

Tests failing:

1. •Keep task in-progress
2. •Fix the issue
3. •Re-run verification

Forgot to stop timer: {{#if mcp}}

mcp__knowns__add_time({
  "taskId": "$ARGUMENTS",
  "duration": "<duration>",
  "note": "Timer correction"
})

{{else}}

knowns time add $ARGUMENTS <duration> -n "Timer correction"

{{/if}}

Remember

• •Check AC only AFTER work is done
• •Use templates when available
• •Track progress with notes
• •Ask before scope changes
• •Follow the approved plan
• •Verify before marking done
• •Always stop the timer
• •Consider knowledge extraction