Intent Integrity Kit Implement

Execute the implementation plan by processing all tasks in tasks.md.

User Input

$ARGUMENTS

You MUST consider the user input before proceeding (if not empty).

Windows: Replace bash …/iikit-07-implement/scripts/bash/*.sh with pwsh …/iikit-07-implement/scripts/powershell/*.ps1 (same flags, -PascalCase).

Constitution Loading

Load constitution per constitution-loading.md (enforcement mode — extract rules, declare hard gate, validate before every file write).

Prerequisites Check

1. Run: bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/check-prerequisites.sh --phase 07 --json
2. Parse for FEATURE_DIR and AVAILABLE_DOCS. If missing tasks.md: ERROR.
3. If JSON contains needs_selection: true: present the features array as a numbered table (name and stage columns). Follow the options presentation pattern in conversation-guide.md. After user selects, run:

   bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/set-active-feature.sh --json <selection>

   Then re-run the prerequisites check from step 1.

Pre-Implementation Validation

Bugfix detection: if every unchecked task has a T-B prefix, this is a bugfix-only run — relaxed gates below.

Standard mode: Verify artifact completeness (constitution, spec, plan, tasks, checklists), cross-artifact consistency (FR-XXX → tasks, tech stack → file paths), and checklist completion (all 100% or ask user). Report READY or BLOCKED.

Bugfix mode: Only require tasks.md (T-B tasks) and bugs.md (matching BUG-NNN). Skip spec/plan/checklist gates. BDD chain (§2.1–2.4) still applies if .feature files exist.

Dashboard

Suggest the user open the dashboard to watch implementation progress in real time:

Dashboard: file://$(pwd)/.specify/dashboard.html (resolve the path) — updates live as tasks complete

Execution Flow

1. Load Context

Read tasks.md + plan.md (standard) or tasks.md + bugs.md (bugfix). Optional: data-model.md, contracts/, research.md, quickstart.md, tests/features/.

2. TDD Support Check

If tests/features/ directory exists (contains .feature files), verify assertion integrity:

bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/testify-tdd.sh comprehensive-check "FEATURE_DIR/tests/features" "CONSTITUTION.md"

Parse JSON response: PASS (proceed), BLOCKED (halt, show remediation), WARN (proceed with caution).

If TDD mandatory but tests/features/ missing or empty: ERROR with Run: /iikit-04-testify.

2.1 BDD Verification Chain Enforcement

When .feature files exist, the full BDD verification chain applies to each implementation task:

Step 1 — Write step definitions: Write step definition code that binds Gherkin steps to application calls. Place in tests/step_definitions/.

Step 2 — Verify step coverage: All .feature steps must have matching step definitions.

bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/verify-steps.sh --json "FEATURE_DIR/tests/features" "FEATURE_DIR/plan.md"

Must return PASS before continuing. If BLOCKED: fix missing step definitions. If DEGRADED: proceed with caution (no BDD framework available).

Step 3 — RED phase: Run the BDD tests. They MUST fail (step definitions exist but production code does not yet implement the behavior). This confirms the tests are meaningful.

Step 4 — Write production code: Implement the feature code that makes the tests pass.

Step 5 — GREEN phase: Run the BDD tests again. They MUST pass. If they fail: fix the production code, not the tests or .feature files.

Step 6 — Verify step quality: Ensure step definitions have meaningful assertions (not empty bodies or tautologies).

bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/verify-step-quality.sh --json "FEATURE_DIR/tests/step_definitions" "<language>"

Must return PASS before marking the task complete. If BLOCKED: fix the flagged step definitions.

2.2 Feature File Immutability

CRITICAL: .feature files MUST NOT be modified during implementation. They are generated by /iikit-04-testify and hash-locked. Only step definitions and production code may be modified. If a .feature file needs changes, re-run /iikit-04-testify.

2.3 Test Execution Enforcement

Tests MUST be run, not just written. After writing a test: run it immediately (expect red). After implementing: run it (expect green). If tests fail: fix code, not tests. Never mark a test task [x] without execution output.

bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/verify-test-execution.sh verify "FEATURE_DIR/tests/features" "$(cat test-output.log)"

Block on any status other than PASS.

2.4 Task Completion Gate

A task is NOT complete until:

1. verify-steps.sh returns PASS (all steps defined)
2. BDD tests pass (GREEN phase confirmed)
3. verify-step-quality.sh returns PASS (no empty/trivial assertions)

Do NOT mark [x] in tasks.md until all three gates pass.

3. Setup (Dependencies, Tiles, Scaffolding)

Before writing source code:

1. Install dependencies from plan.md Technical Context (detect package manager, add runtime + dev deps, commit manifest + lockfile)
2. Install Tessl tiles for each major dependency: tessl search <pkg> then tessl install <tile>. Query tile docs before writing library code. See tessl-integration.md.
3. Scaffold project if needed. For existing directories, use force/overwrite flags. See ignore-patterns.md for gitignore patterns.

6. Parse and Execute Tasks

6.1 Task extraction: parse tasks.md for phase, completion status ([x] = skip), dependencies, [P] markers, [USn] labels. Build in-memory task graph.

6.2 Execution strategy — read parallel-execution.md BEFORE proceeding: If tasks.md contains [P] markers, you MUST use the Task tool to dispatch parallel batches as concurrent subagents (one worker per task). Only fall back to sequential execution if the runtime has no subagent dispatch mechanism. Report mode per formatting-guide.md (Execution Mode Header).

6.3 Phase-by-phase:

1. Collect eligible tasks (dependencies satisfied)
2. Build parallel batches from [P] tasks with no mutual dependencies
3. Dispatch — parallel: launch one Task tool subagent per [P] task in the batch; sequential: one at a time
4. Collect results, checkpoint [x] in tasks.md per batch, then commit per task (§6.6)
5. Repeat until phase complete

Cross-story parallelism: independent stories can run as parallel workstreams after Phase 2 (verify no shared file modifications).

6.4 Rules: install dependencies (§3) and Tessl tiles (§4) before writing code, query tiles before library code, tests before code if TDD, run tests after writing them, only orchestrator updates tasks.md.

6.5 Failure handling: let in-flight siblings finish, mark successes, report failures, halt phase. Constitutional violations in workers: worker stops, reports to orchestrator, treated as task failure.

6.6 Task Commits: After each task is marked [x], stage its changed files (git add specific files, NOT -A) and commit:

• <feature-id> = FEATURE_DIR with specs/ prefix and trailing / stripped (e.g. 001-user-auth)
• Subject: feat(<feature-id>): <task-id> <task description> (use fix(…) for T-B tasks)
• Trailers: iikit-feature: <feature-id> and iikit-task: <task-id>
• Skip if no files changed; for parallel batches commit each task individually after batch completes
• After each commit, regenerate the dashboard so the board reflects the latest task state:

  bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/generate-dashboard-safe.sh

7. Output Validation

Before writing ANY file: review against constitutional principles. On violation: STOP, explain, suggest alternative.

8. Progress Tracking

Report after each task/batch. Mark completed [x] in tasks.md. Halt on failure.

9. Post-Fix GitHub Integration (Bug Fix Tasks)

After completing bug fix tasks (tasks with T-B prefix pattern):

1. Check if FEATURE_DIR/bugs.md exists
2. For each completed bug (all T-BNNN tasks for a BUG-NNN marked [x]):
   • Read the GitHub Issue field from the bug's entry in bugs.md
   • If a GitHub issue is linked (e.g., #42):
     • Close via commit: include Fixes #<number> in the last task's commit message (§6.6) — GitHub auto-closes the issue when pushed/merged
     • Post a comment: use gh issue comment if available, otherwise curl the GitHub API (POST /repos/{owner}/{repo}/issues/{number}/comments). Comment content: root cause from bugs.md, completed fix tasks, and fix reference
   • If no GitHub issue is linked: skip silently

10. Completion

All tasks [x], features validated against spec, test execution enforcement (§2.1) satisfied, Tessl usage reported.

Error Handling

Missing artifacts: STOP with run instructions. Constitution violations: STOP, explain, suggest alternative. Checklist incomplete: ask user. Task/parallel failure: report + halt (§6.5). Tests not run: STOP. Tests failing: fix code, re-run.

Next Steps

Run: bash .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/bash/next-step.sh --phase 07 --json Windows: pwsh .tessl/tiles/tessl-labs/intent-integrity-kit/skills/iikit-07-implement/scripts/powershell/next-step.ps1 -Phase 07 -Json

Parse the JSON and present:

1. If clear_after is true: suggest /clear before proceeding
2. If next_step is /iikit-07-implement (feature incomplete): suggest resuming implementation
3. If next_step is null (feature complete): congratulate and list alt_steps
4. If alt_steps non-empty: list as alternatives (e.g., /iikit-08-taskstoissues)
5. For next_step and each alt_step, include the model_tier from the JSON so the user knows which model is best for each option. Look up tiers in model-recommendations.md for agent-specific switch commands.
6. Append dashboard link

Push commits to remote if available: git push. If on a feature branch, offer to merge. Ask the user which approach they prefer:

• A) Merge locally: git checkout main && git merge <branch>
• B) Create PR: gh pr create if available, otherwise provide the GitHub URL to create one manually
• C) Skip: user will handle it

Format:

Implementation complete!
Next: [/clear → ] <next_step or "All tasks done!">
[- <alt_step> — <reason> (model: <tier>)]

- Dashboard: file://$(pwd)/.specify/dashboard.html (resolve the path)