home / skills / duc01226 / easyplatform / cook-hard

cook-hard skill

/.claude/skills/cook-hard

This skill guides rigorous implementation with maximum verification, ensuring code correctness, security, and thorough reviews for critical production features.

npx playbooks add skill duc01226/easyplatform --skill cook-hard

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
3.0 KB
---
name: cook-hard
description: "[Implementation] ⚡⚡⚡⚡ Thorough implementation with maximum verification"
argument-hint: [tasks]
infer: true
---

**Ultrathink** to plan and implement these tasks with maximum verification:
<tasks>$ARGUMENTS</tasks>

**Mode:** HARD - Extra research, detailed planning, mandatory reviews.

## ⚠️ MUST READ Before Starting

**IMPORTANT: You MUST read these shared protocols before any code changes. Do NOT skip.**

- **⚠️ MUST READ** `.claude/skills/shared/anti-hallucination-protocol.md` — Assumption validation, evidence chains, context anchoring
- **⚠️ MUST READ** `.claude/skills/shared/knowledge-graph-template.md` — Per-file analysis structure (for investigation phases)

**Core Rule:** Verify every assumption with actual code evidence before making changes. Search for all usages, read implementations, trace dependencies. If confidence < 90%, investigate further or ask user.

## Workflow

### 1. Deep Research Phase
- Launch 2-3 `researcher` subagents in parallel covering:
  - Technical approach validation
  - Edge cases and failure modes
  - Security implications
  - Performance considerations
- Use `/scout-ext` for comprehensive codebase analysis
- Generate research reports (max 150 lines each)

### 2. Comprehensive Planning
- Use `planner` subagent with all research reports
- Create full plan directory with:
  - `plan.md` - Overview with risk assessment
  - `phase-XX-*.md` - Detailed phase files
  - Success criteria for each phase
  - Rollback strategy

### 3. Verified Implementation
- Implement one phase at a time
- After each phase:
  - Run type-check and compile
  - Run relevant tests
  - Self-review before proceeding

### 4. Mandatory Testing
- Use `tester` subagent for full test coverage
- Write tests for:
  - Happy path scenarios
  - Edge cases from research
  - Error handling paths
- NO mocks or fake data allowed
- Repeat until all tests pass

### 5. Mandatory Code Review
- Use `code-reviewer` subagent
- Address all critical and major findings
- Re-run tests after fixes
- Repeat until approved

### 6. Documentation Update
- Use `docs-manager` to update relevant docs
- Use `project-manager` to update project status
- Record any architectural decisions

### 7. Final Report
- Summary of all changes
- Test coverage metrics
- Security considerations addressed
- Unresolved questions (if any)
- Ask user to review and approve

## When to Use

- Critical production features
- Security-sensitive changes
- Public API modifications
- Database schema changes
- Cross-service integrations

## Quality Gates

| Gate     | Criteria                  |
| -------- | ------------------------- |
| Research | 2+ researcher reports     |
| Planning | Full plan directory       |
| Tests    | All pass, no mocks        |
| Review   | 0 critical/major findings |
| Docs     | Updated if needed         |

## IMPORTANT Task Planning Notes

- Always plan and break many small todo tasks
- Always add a final review todo task to review the works done at the end to find any fix or enhancement needed

Overview

This skill codifies a rigorous, high-assurance implementation workflow for critical C# changes. It enforces multi-agent research, staged planning, verified implementation, exhaustive testing without mocks, mandatory code review, and complete documentation and reporting. The goal is maximum confidence (>=90%) before any production change is accepted.

How this skill works

The skill runs parallel researcher subagents to validate approaches, enumerate edge cases, and collect evidence chains. A planner then synthesizes those reports into a phased plan with success criteria and rollback strategy. Each implementation phase is compiled and tested, followed by independent test generation, a no-mocks policy, and iterative code review until no critical or major findings remain. Finally, docs and a final report are produced for user approval.

When to use it

  • Making changes to critical production features
  • Implementing security-sensitive fixes or hardening
  • Modifying public APIs or backward-incompatible endpoints
  • Performing database schema migrations or data model changes
  • Coordinating cross-service integrations or architectural shifts

Best practices

  • Always validate assumptions with code-level evidence and usage searches before changing code
  • Run at least two independent researcher reports covering technical, security, and performance angles
  • Break work into small, verifiable phases with explicit success criteria and rollback steps
  • Write tests that exercise real behavior and data; avoid mocks and synthetic shortcuts
  • Require an independent code-reviewer and re-run tests after each remediation

Example use cases

  • Refactoring a core authentication flow that affects multiple services
  • Adding encryption-at-rest for sensitive database fields with migration steps
  • Changing a public HTTP API contract used by external clients
  • Optimizing a hot path with potential performance and correctness trade-offs
  • Introducing a cross-service event schema that requires coordinated deploys

FAQ

What evidence is required before making a code change?

Trace usages, run searches for all call sites, inspect implementations and tests, and collect concrete examples that confirm or refute assumptions. If confidence is under 90%, continue research.

Are mocks allowed in tests?

No — tests must exercise real components and data where feasible. If integration is impossible in CI, document why and provide mitigations and additional validation steps.

How many researcher reports are mandatory?

At least two independent researcher reports covering technical approach, failure modes, security, and performance; combine findings into the planner.