implement

name	implement
description	Generate production code and tests for a story group using agent teams for parallel execution.
argument-hint	[group-id]

Implement Skill

Generate production-quality code and tests for all stories in a dependency group, using a Claude Code agent team for parallel execution.

---

Usage

/implement C

Implements all stories in group C. The group ID corresponds to a node in specs/stories/dependency-graph.md.

---

Prerequisites

Before running /implement, verify:

• specs/stories/dependency-graph.md exists and lists groups with story assignments.
• specs/design/component-map.md exists and maps each story to the files it owns.
• All stories in the target group have acceptance criteria written.
• All upstream groups are already implemented and passing evaluation.

If any prerequisite is missing, stop and report what is absent. Do not proceed with partial context.

---

Execution Steps

Step 1 — Load Quality Principles

Read .claude/skills/code-gen/SKILL.md in full. These six principles (small modules, static typing, functions under 50 lines, explicit error handling, no dead code, self-documenting names) apply to every line of code produced. Inject the full text into every teammate prompt.

Step 2 — Load Dependency Graph

Read specs/stories/dependency-graph.md. Identify:

• Which stories belong to the requested group.
• Which groups must be complete before this group (upstream dependencies).
• The total story count for this group.

Abort if upstream groups are not yet evaluated as PASS.

Step 3 — Load Component Map

Read specs/design/component-map.md. For each story in the group, extract:

• The list of files the story owns (may create or modify).
• Any shared interface or type files that multiple stories reference.

This ownership map is the single source of truth for file assignments during parallel execution.

Step 4 — Load Learned Rules

Read .claude/state/learned-rules.md. Inject ALL rules verbatim into every teammate spawn prompt. Learned rules represent project-specific decisions made during previous sprints (naming conventions, library choices, API patterns). Skipping this step causes regressions.

Step 4.5 — Read Model Routing

Read project-manifest.json field execution.model_routing. This determines which model backs each agent spawn:

• cloud-only: Spawn agents normally (Claude Code defaults).
• hybrid: Generator teammates use the code_gen_agents.base_url endpoint. If using an OpenAI-compatible local model, prefix agent prompts with: Use model: {model_name} via {base_url}.
• local-only: ALL agent spawns (teammates, evaluator, reviewers) use the local endpoint.

If strategy is hybrid or local-only, verify the local endpoint is reachable before spawning:

curl -sf {base_url}/models > /dev/null || echo "WARN: Local model not reachable"

If unreachable, warn the human and wait for instructions. Do not fall back silently.

Step 5 — Spawn Agent Team (Multiple Stories)

If the group contains 2 or more stories, spawn a Claude Code agent team:

• Create 1 teammate per story, up to a maximum of 5 concurrent teammates.

• If the group has more than 5 stories, batch them: first 5 stories run, then the remainder after all complete.

• Each teammate spawn prompt must include:

  • The story's acceptance criteria (full text).
  • The file ownership list from component-map.md for that story.
  • All learned rules from .claude/state/learned-rules.md.
  • All six quality principles from .claude/skills/code-gen/SKILL.md.
  • Instruction to message teammates before modifying any shared type or interface file.
  • Instruction to await plan approval before writing any code (present the plan, wait for confirmation).

• Teammates must coordinate on shared types:

  • Before editing a type definition used by another story's files, send a message to the affected teammate describing the change.
  • Teammate receiving the message must acknowledge before the edit proceeds.

• If two teammates claim ownership of the same file, escalate to the orchestrator (this agent). Do not merge partial changes. Resolve ownership, then continue.

Step 6 — Use Generator Directly (Single Story)

If the group contains exactly 1 story, do not spawn a team. Execute the story using the generator agent directly:

• Present a plan (files to create/modify, type definitions, test strategy).
• Await approval.
• Implement code, then write tests.

Step 7 — Validation Gate

After all teammates (or the generator) complete:

1. Run the full test suite: npm test or pytest (whichever applies to the project).
2. Run the linter: npm run lint or ruff check ..
3. Run the type checker: tsc --noEmit or mypy ..

All three must pass with zero errors before proceeding. If any fails, return the failure output to the responsible teammate for a fix, then re-run the validation gate.

Step 8 — Code Review

Spawn the code-reviewer agent on the set of changed files:

• Pass the list of modified files and the story acceptance criteria.
• The reviewer emits findings at three severity levels: BLOCK, WARN, INFO.
• BLOCK findings must be fixed. Spawn the responsible teammate to address the finding, re-run tests, re-run the reviewer. Maximum 3 retry cycles.
• WARN findings are logged but do not block merge.
• INFO findings are optional improvements.

If the reviewer still emits BLOCK findings after 3 retries, escalate to the user with a summary of the unresolved issues.

---

Rules

• Every file produced must trace to a story in the current group. No story, no code.
• Code is written first; tests are written after, against the public interface.
• No speculative code ("might need later"). If it is not in an acceptance criterion, it does not exist.
• Teammates may not edit files outside their ownership assignment without coordinator approval.
• Plan approval is mandatory before any teammate begins coding. Skipping this step is not a time-saver — it causes conflicts and rework.

---

Gotchas

• Teammates editing the same file: Prevent this with the ownership map. If it happens anyway, stop both teammates, resolve ownership, reconcile changes manually.
• Skipping plan approval: Leads to scope creep, missed acceptance criteria, and merge conflicts. Always require the plan step.
• Deferring test coverage: Tests are written in the same sprint cycle, not later. "I'll add tests in the next sprint" is not acceptable.
• Vibe coding without acceptance criteria: Every function must trace to an acceptance criterion. If the criterion does not exist, do not write the code — write the criterion first.
• Ignoring learned rules: Failing to inject .claude/state/learned-rules.md recreates decisions the team has already made, causing style and pattern drift.