From tac
Generates and refines detailed plans from /chore, /bug, /feature templates. Ensures specificity, completeness, testability via codebase analysis and validation steps.
npx claudepluginhub melodic-software/claude-code-plugins --plugin tacThis skill is limited to using the following tools:
Guide for generating high-quality plans from meta-prompt templates.
Creates specs and phased file-level implementation plans for features, bugs, refactors by researching codebase with search, graph queries, and context files.
Guides writing or updating detailed plans for Claude Code sessions, assessing component readiness, filling knowledge gaps, defining testing strategies, and verifying hypotheses.
Generates implementation plans using minimal, standard, or comprehensive templates matched to task complexity. Use for planning bugs, features, changes, or architectural work.
Share bugs, ideas, or general feedback.
Guide for generating high-quality plans from meta-prompt templates.
/chore, /bug, /feature templatesParse the high-level description:
Input: "Add user authentication with OAuth"
Extract:
- Domain: Authentication
- Scope: User-facing, OAuth protocol
- Complexity: Medium-high (external integration)
Use extended thinking for complex planning:
THINK HARD about:
- What files need to change?
- What are the dependencies?
- What could go wrong?
- How will we verify success?
Gather context before planning:
Complete every section with specifics:
## Relevant Files
- src/auth/OAuthProvider.ts (create)
- src/auth/index.ts (modify - add export)
- src/config/oauth.ts (create)
- tests/auth/oauth.test.ts (create)
Check against quality criteria before finalizing.
Every plan should meet these standards:
Bad: "Update the component" Good: "Update UserProfile.tsx to add loading state on line 45"
Bad: "Handle errors appropriately" Good: "Add try/catch in fetchUser(), log errors with console.error, show ErrorBoundary"
All template sections filled:
Every plan must include validation:
## Validation Commands
- Run `npm test -- auth` to verify unit tests pass
- Run `npm run e2e -- oauth` to verify integration
- Manual: Complete OAuth flow in browser
Focus on:
Focus on:
Focus on:
Problem: "Implement the feature" Fix: Break into specific sub-tasks with file references
## Step by Step Tasks
1. Create AuthContext in src/contexts/AuthContext.tsx
2. Add useAuth hook in src/hooks/useAuth.ts
3. Wrap App component with AuthProvider in src/App.tsx
4. Add login route in src/routes/index.tsx
Problem: Plan doesn't mention test files Fix: Always include test file creation/modification
## Relevant Files
- src/components/Login.tsx (create)
- src/components/Login.test.tsx (create) # Tests!
Problem: "Make sure it works" Fix: Specific commands with expected outcomes
## Validation Commands
- Run `npm test` - expect 0 failures
- Run `npm run build` - expect successful build
- Run `npm run lint` - expect 0 errors
Problem: Plan addresses more than requested Fix: Stay focused on original request, note related work in Notes
## Notes
- Related: Login form could use accessibility improvements (separate chore)
- Related: Password reset flow needs similar OAuth option (separate feature)
Note: Extended thinking is enabled by default in Claude Code (31,999 tokens). Phrases like "think", "think hard", "ultrathink" are prompt instructions that signal your expectations but don't control token allocation. Use Alt+T to toggle thinking or MAX_THINKING_TOKENS to adjust budget.
Use "THINK HARD" in meta-prompts as a prompt instruction to signal the need for careful reasoning:
| Phrase | When to Use |
|---|---|
| "THINK HARD" | Medium-to-high complexity, encourage careful reasoning |
specs/[type]-[descriptive-name].md
Examples:
- specs/chore-update-dependencies.md
- specs/bug-fix-login-race-condition.md
- specs/feature-oauth-authentication.md
Always write plans to specs/ directory (or project-specific equivalent).
Match the template's Plan Format exactly - don't add or remove sections.
Date: 2025-12-26 Model: claude-opus-4-5-20251101