---

name: openagent-skill-index description: Use when starting any conversation - establishes OpenAgent skills and approval gate philosophy

OpenAgent Skill Index

Core principle: OpenAgent skills integrate approval gates throughout proven development workflows, ensuring safety-first development.

Skill Catalog

When working on coding tasks, use these OpenAgent custom skills:

Task Type	Use This Skill
Design/brainstorming new features	custom/openagent-brainstorming
Implementing code (any feature/bugfix)	custom/openagent-test-driven-development
Debugging/troubleshooting	custom/openagent-systematic-debugging
Creating isolated workspace	custom/openagent-using-git-worktrees
Writing implementation plans	custom/openagent-writing-plans
Executing written plans	custom/openagent-executing-plans
Subagent-driven development	custom/openagent-subagent-driven-development
Finishing development branch	custom/openagent-finishing-a-development-branch
Requesting code review	custom/openagent-requesting-code-review
Receiving code review feedback	custom/openagent-receiving-code-review
Verification before completion	custom/openagent-verification-before-completion
Dispatching parallel agents	custom/openagent-dispatching-parallel-agents
Creating new skills	custom/openagent-writing-skills
Understanding skill usage	custom/openagent-using-skills

OpenAgent Philosophy

OpenAgent skills integrate approval gates at phase transitions:

Approval gate pattern:

• ⏸️ Request approval before major operations
• ⏸️ Report results after operations complete
• ⏸️ Stop on failure, propose fixes, request approval
• ⏸️ Ask permission for destructive actions

Why approval gates matter:

• Prevents accidental destructive operations
• Ensures user awareness at each phase
• Creates natural checkpoints for review
• Aligns with safety-first development

Automatic Skill Detection

When you see these user requests:

User Says	You Should Think	Load This Skill
"Build/add/create [feature]"	Design needed	custom/openagent-brainstorming
"Write a plan for..."	Planning needed	custom/openagent-writing-plans
"Execute this plan"	Batch execution	custom/openagent-executing-plans
"Implement [feature]"	TDD implementation	custom/openagent-test-driven-development
"Fix this bug"	Systematic debugging	custom/openagent-systematic-debugging
"Start working on [feature]"	Need isolated workspace	custom/openagent-using-git-worktrees
"Write tests for..."	TDD workflow	custom/openagent-test-driven-development
"Something's not working"	Debug systematically	custom/openagent-systematic-debugging
"Review this code"	Request review	custom/openagent-requesting-code-review
"I'm done with this feature"	Finish branch	custom/openagent-finishing-a-development-branch
"Verify this works"	Verification needed	custom/openagent-verification-before-completion
"Multiple independent failures"	Parallel dispatch	custom/openagent-dispatching-parallel-agents

Workflow Integration

Typical OpenAgent workflow:

1. New Feature Request

   User: "Add user authentication"
   
   You: Load custom/openagent-brainstorming
        → Request approval to start brainstorming
        → Ask questions, explore design
        → Request approval to save design doc
        → Request approval to create worktree (load custom/openagent-using-git-worktrees)
        → Request approval to begin TDD (load custom/openagent-test-driven-development)
   

2. Bug Report

   User: "Login endpoint returning 500 error"
   
   You: Load custom/openagent-systematic-debugging
        → Request approval to investigate
        → Phase 1: Root cause investigation
        → Phase 2: Pattern analysis
        → Phase 3: Hypothesis testing
        → Phase 4: Create test + fix (use custom/openagent-test-driven-development)
   

3. Implementation Task

   User: "Implement the user registration endpoint"
   
   You: Load custom/openagent-test-driven-development
        → Request approval to start TDD
        → RED: Write failing test
        → Verify test fails
        → Request approval to implement
        → GREEN: Write minimal code
        → Verify tests pass
        → Request approval to commit
   

Checking for Skills

Before ANY response (even clarifying questions), check:

1. Is this a new feature/design? → Load custom/openagent-brainstorming
2. Is this implementation/code change? → Load custom/openagent-test-driven-development
3. Is this a bug/failure/unexpected behavior? → Load custom/openagent-systematic-debugging
4. Does this need isolated workspace? → Load custom/openagent-using-git-worktrees
5. Writing or executing implementation plans? → Load custom/openagent-writing-plans or custom/openagent-executing-plans
6. Multiple independent tasks/failures? → Load custom/openagent-subagent-driven-development or custom/openagent-dispatching-parallel-agents
7. Code review needed? → Load custom/openagent-requesting-code-review or custom/openagent-receiving-code-review
8. About to claim completion? → Load custom/openagent-verification-before-completion
9. Development work complete? → Load custom/openagent-finishing-a-development-branch

The rule from openagent-using-skills applies:

"Invoke relevant or requested skills BEFORE any response or action. Even a 1% chance a skill might apply means that you should invoke the skill to check."

All OpenAgent Custom Skills Available

All 14 OpenAgent skills with approval-gated workflows:

Tier 1: Core Development Workflows (8 skills)

• custom/openagent-test-driven-development - TDD with approval gates
• custom/openagent-systematic-debugging - 4-phase debugging with approval gates
• custom/openagent-brainstorming - Design exploration with approval gates
• custom/openagent-using-git-worktrees - Isolated workspace with approval gates
• custom/openagent-executing-plans - Plan execution with approval gates
• custom/openagent-writing-plans - Implementation planning with approval gates
• custom/openagent-subagent-driven-development - Subagent workflows with approval gates
• custom/openagent-finishing-a-development-branch - Branch completion with approval gates

Tier 2: Code Quality & Review (4 skills)

• custom/openagent-requesting-code-review - Code review requests with approval gates Tier 2: Code Quality & Review (4 skills)
• custom/openagent-requesting-code-review - Code review requests with approval gates
• custom/openagent-receiving-code-review - Review feedback handling with approval gates
• custom/openagent-verification-before-completion - Evidence-based completion with approval gates
• custom/openagent-dispatching-parallel-agents - Parallel task dispatch with approval gates

Foundation Skills (2 skills)

• custom/openagent-using-skills - Skill usage methodology
• custom/openagent-writing-skills - TDD for creating new skills

Red Flags - Skill Not Loaded

If you catch yourself:

• Writing code without loading custom/openagent-test-driven-development
• Starting to debug without loading custom/openagent-systematic-debugging
• Asking design questions without loading custom/openagent-brainstorming
• Creating worktree without loading custom/openagent-using-git-worktrees
• Writing/executing plans without loading the appropriate skill
• Requesting/receiving code review without loading the appropriate skill
• About to claim completion without loading custom/openagent-verification-before-completion
• Finishing branch without loading custom/openagent-finishing-a-development-branch

STOP and load the appropriate skill first.

Summary

The key principle:

OpenAgent Skills = Proven Methodology + Approval Gates

Your job:

1. Detect task type from user request
2. Load appropriate custom/openagent-* skill
3. Follow skill workflow with approval gates
4. Request approval before major operations