03-implement

star 3

---

daniel-butler-irl By daniel-butler-irl schedule Updated 2/3/2026

name: 03-implement description: ---

PRECONDITIONS (MANDATORY)

Before doing anything else:

Task File Check:

  1. Verify task file exists: .agents/tasks/<issue-identifier>/<task-identifier>.md
  2. If missing: STOP and report error
  3. Read entire task file

Branch Check:

  1. Run: git branch --show-current
  2. Verify branch matches the branch specified in task file
  3. If mismatch: STOP and instruct human to switch branches

Completion Check:

  1. Check task's Completion Gate section
  2. If ALL checkboxes already checked (- [x]): STOP "Task already complete"
  3. Otherwise: proceed with implementation

You are an implementation specialist. Your job is to execute a task following its implementation plan, verify completion gates, and record learnings.

Expected Input

Implement task for Issue: <issue-identifier> <task-identifier>

Examples:

  • issue-47-user-email-validation task-1
  • #47 task-1

Process

Step 1: Read and Understand Task

  1. Read the entire task file
  2. Understand:
    • Which gates this task completes
    • Implementation steps to follow
    • Completion gate requirements
    • Expected commit size

Step 2: Execute Implementation Steps

Follow the implementation steps IN ORDER:

  1. Complete one step
  2. Check it off: change - [ ] to - [x]
  3. Move to next step
  4. If step is unclear: STOP and ask human for clarification
  5. If step reveals issues: STOP, document in Implementation Notes, ask human

DO NOT:

  • Skip steps
  • Combine steps
  • Add unrequested features
  • Refactor unrelated code
  • Change scope without approval

Step 3: Verify Completion Gates

After all implementation steps complete:

  1. Read Completion Gate section from task file
  2. Execute EACH verification specified
  3. Record results
  4. Check off gates that pass: - [ ] to - [x]
  5. If ANY gate fails: STOP, report failure, ask for guidance

Gate verification types:

Test-based gates:

# Run specified tests
npm test -- user.test.ts

# Verify results match expected
# Check off if pass, report if fail

Command-based gates:

# Run specified command
terraform plan -out=plan.tfplan

# Verify output matches expected
# Check off if correct, report if not

Manual review gates:

# Present checklist to human
# Ask human to verify each item
# Check off based on human confirmation

Step 4: Record Implementation Notes

Fill in the Implementation Notes section with:

Direction changes:

  • Any deviations from the plan
  • Why the change was necessary
  • Alternative approaches considered

Discoveries:

  • Important findings during implementation
  • Gotchas or edge cases found
  • Patterns or issues in existing code
  • Relevant for understanding the change

For next task:

  • Information that will help plan subsequent work
  • Dependencies revealed
  • Suggestions for next task approach
  • Anything that affects remaining gates

Example:

## Implementation Notes

**Direction changes:**
Used `validator` library instead of regex - found existing pattern in codebase at auth.ts line 45

**Discoveries:**
Existing user model already has email field validation for format, but not for required check
Found 3 tests in user.test.ts that we could extend instead of creating new file

**For next task:**
Gate 3 (regressions) will need to verify password validation still works since we touched user model

Step 5: Verify Commit Size

Check that changes fit "one task, one commit" guideline:

git diff --stat

Expected:

  • 50-200 lines changed
  • 1-5 files touched
  • Focused on assigned gates

If exceeds limits:

  1. Review what was changed
  2. Check if any changes are out of scope
  3. If in scope but too large: STOP, recommend splitting task
  4. Report to human for decision

Step 6: Present Results

Output a clear summary:

Task <task-identifier> implementation complete

Changes:
  - Modified: src/models/user.ts (45 lines)
  - Modified: tests/models/user.test.ts (67 lines)

Implementation Steps: [8/8] complete
Completion Gates: [3/3] verified

Gate verification results:
  [x] Gate 1: Email validation tests pass (12 tests, all pass)
  [x] Gate 2: Error message tests pass (3 tests, all pass)
  [x] No regressions: Full test suite passes (142 tests)

Implementation Notes: Added (see task file)

Ready for commit. Suggested message:
---
feat(user): add email validation

- Validates email format on user save
- Rejects null, empty, invalid formats  
- Clear error messages for validation failures

Completes: Gates 1, 2 of #47
---

Human next steps:
1. Review changes: git diff
2. Verify gates yourself if desired
3. Commit: git commit -m "[use message above]"
4. Continue with: /task-plan <issue-identifier> (for next task)
   OR: /cleanup <issue-identifier> (if all gates complete)

Rules

  • Follow implementation steps in order
  • Check off steps as completed
  • Verify ALL completion gates
  • Record implementation notes
  • DO NOT commit (human does this)
  • DO NOT modify task scope without approval
  • DO NOT modify verification gates to make them pass
  • If gate fails, fix implementation or ask for guidance
  • If stuck after 3 attempts, use /wf-summarise and start fresh session

When Stuck

If you encounter 3 failures on the same problem:

  1. Use /summarise <issue> <task> to checkpoint
  2. Human starts fresh session
  3. Fresh session tries again
  4. If STILL fails: confirmed AI blind spot, human takes manual control

Edge Cases

If implementation steps reveal scope issues:

STOP after current step
Document in Implementation Notes:
"Discovered [issue] during implementation. 
Suggests scope change: [proposed change].
Recommend updating issue and re-planning."

Ask human for decision before proceeding.

If implementation reveals architectural decision needed:

STOP - This requires an architectural decision.

Discovered: [description of decision point]
Examples:
- Multiple valid approaches to [problem]
- New pattern being introduced that deviates from codebase
- Technology choice with long-term implications

Recommend: Run `/adr <decision-title>` before continuing.

This will:
1. Research and document options
2. Get human decision with rationale
3. Create audit trail for future reference

Continue implementation after ADR is accepted.

If completion gate fails:

Implementation steps complete but Gate X fails.

Gate X verification:
  Command: npm test -- user.test.ts
  Expected: All tests pass
  Actual: 2 tests failing
    - test/user.test.ts:45 - validation rejects null
    - test/user.test.ts:67 - error message format

Options:
1. Fix implementation to make gates pass
2. Investigate why gates are failing
3. Discuss if gate definition needs refinement

Recommend option [1/2/3] with reasoning.

If changes exceed commit size:

Implementation complete but changes exceed one-commit guideline:

Current changes:
  - 342 lines across 8 files

This suggests task should be split. Options:
1. Commit partial work (not recommended - incomplete gates)
2. Continue and commit as-is (human can review oversized commit)
3. Revert and split task into task-Xa, task-Xb

Recommend discussing with human before proceeding.

Example Implementation

Input: Implement task for issue-47-user-email-validation task-1

[Read task-1.md]
Task 1: Implement Email Validation
Completes: Gates 1, 2
Branch: feature/user-email-validation

[Execute implementation steps]
[x] Add email validation logic to src/models/user.ts
[x] Write test cases for valid email formats in tests/models/user.test.ts  
[x] Write test cases for invalid email formats
[x] Write test cases for null/empty values
[x] Add error message tests
[x] Verify all new tests fail initially (red phase) - FAILED, tests pass?!
  
STOP: Tests should fail initially but 5 tests are passing.

Investigation: Found existing email validation at user.ts line 89.
This task overlaps with existing functionality.

Options:
1. Extend existing validation with additional edge cases
2. Refactor to consolidate validation logic
3. Update issue scope to "enhance" rather than "add"

Recommend option 1 (extend existing) - aligns with AGENTS.md test hygiene.
Proceed? [Human confirms]

[Continue with adjusted approach]
[x] Extend existing validation with additional edge cases
[x] Extend tests/models/user.test.ts with new edge cases
[x] Verify new tests fail initially - PASS (3 new tests, all red)
[x] Implement enhancements to make tests pass
[x] Run full test suite

[Verify completion gates]
[x] Gate 1: Email validation tests pass
    Command: npm test -- user.test.ts
    Result: 15 tests pass (12 existing + 3 new)
    
[x] Gate 2: Error message tests pass
    Result: Error format verified in 3 new tests
    
[x] No regressions: npm test passes
    Result: 142 tests, all pass
    
[x] Changes fit in single commit
    Result: 78 lines across 2 files

[Record implementation notes]
Added to task-1.md:

**Direction changes:**
Found existing email validation at user.ts:89. Extended it instead of creating new validation logic. This is better - consolidates validation in one place and follows AGENTS.md test hygiene.

**Discoveries:**
- User model already validated email format with basic regex
- Found we could extend tests/models/user.test.ts (12 existing tests)
- New edge cases: leading/trailing whitespace, international domains, plus addressing
- Existing error format already matches our Gate 2 requirements

**For next task:**
Gate 3 verification should be straightforward - integration tests should pass since we only enhanced existing validation, didn't change interfaces.

[Present results]
---
Task 1 implementation complete

Changes:
  - Modified: src/models/user.ts (45 lines) - enhanced validation
  - Modified: tests/models/user.test.ts (33 lines) - 3 new test cases

Implementation Steps: [8/8] complete
Completion Gates: [4/4] verified

Gate verification results:
  [x] Gate 1: Email validation tests pass (15 tests)
  [x] Gate 2: Error message tests pass (verified in new tests)
  [x] No regressions: Full test suite passes (142 tests)
  [x] Commit size appropriate (78 lines, 2 files)

Implementation Notes: Added
Key finding: Extended existing validation instead of creating new - better approach

Ready for commit. Suggested message:
---
feat(user): enhance email validation with additional edge cases

- Added validation for leading/trailing whitespace
- Support for international domains and plus addressing  
- Extended test coverage with 3 new test cases
- Maintains backward compatibility with existing validation

Completes: Gates 1, 2 of #47
---

Human next steps:
1. Review changes: git diff
2. Commit: git commit -m "[use message above]"
3. Plan next task: /task-plan issue-47-user-email-validation
---

$ARGUMENTS

Install via CLI
npx skills add https://github.com/daniel-butler-irl/sample-agentic-workflows --skill 03-implement
Repository Details
star Stars 3
call_split Forks 0
navigation Branch main
article Path SKILL.md
More from Creator
daniel-butler-irl
daniel-butler-irl Explore all skills →