Claude-skill-registry controlled-implementation

Execute implementation work in a strictly controlled manner.

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/controlled-implementation" ~/.claude/skills/majiayu000-claude-skill-registry-controlled-implementation && rm -rf "$T"
manifest: skills/data/controlled-implementation/SKILL.md
source content

Controlled Implementation Skill

Execute implementation work in a strictly controlled, step-by-step manner based on approved design and implementation documents, preventing scope creep, premature execution, and task skipping.

Preconditions (Must Read First)

Before starting any implementation, the agent MUST:

  1. Read and fully understand all referenced documents provided by the user (design specs, architecture docs, implementation plans)
  2. Make no assumptions. If anything is unclear or contradictory, stop and ask for clarification

Core Rules (Non-Negotiable)

RuleDescription
No Immediate ImplementationReading and analysis only until explicit permission to start implementation is given
Single-Task ExecutionImplement one task at a time, exactly as defined. Never batch tasks. Never partially start, pre-empt, or prepare future tasks
Mandatory Completion CheckpointAfter completing a task, stop all further work immediately and provide a short, clear completion summary
Explicit Approval GateProceed to the next task only after receiving explicit approval ("green light") from the user
No Forward-Looking WorkDo not anticipate future steps, refactor unrelated areas, or add "helper" changes for upcoming tasks

Expected Behaviour

  • Operate as a senior engineer in a production environment
  • Optimise for safety, clarity, and reviewability
  • Follow provided documents as the single source of truth
  • Prefer correctness over speed

Completion Output Format

After completing each task, respond with:

## ✅ Task Complete: [Task Name or Identifier]

### Summary
- Bullet point 1
- Bullet point 2
- Bullet point 3
- (3–5 bullets describing what was implemented)

### Scope Confirmation
✓ No other tasks were touched
✓ No forward-looking changes made

### Next Step
Awaiting your approval to proceed to: [Next Task Name]

Workflow

flowchart TD
    A[Read Referenced Documents] --> B{Fully Understood?}
    B -->|No| C[Ask for Clarification]
    C --> B
    B -->|Yes| D[Wait for Implementation Permission]
    D --> E[Implement Single Task]
    E --> F[Provide Completion Summary]
    F --> G[Wait for Approval]
    G -->|Approved| H{More Tasks?}
    H -->|Yes| E
    H -->|No| I[Implementation Complete]

Anti-Patterns to Avoid

❌ Anti-Pattern✅ Correct Behaviour
Implementing multiple tasks at onceOne task at a time
Starting work without reading docsRead all docs first
Making "helpful" additionsStrict scope adherence
Preparing code for future tasksCurrent task only
Proceeding without explicit approvalWait for green light
Making assumptions about unclear specsAsk for clarification