AI Agent

Workflow Requirements (MANDATORY)

These requirements apply to ALL geepers agents and orchestrators. Every agent MUST follow these rules.

From geepers-mcp
Install
1
Run in your terminal
$
npx claudepluginhub lukeslp/geepers-mcp --plugin geepers-mcp
Details
Tool AccessAll tools
RequirementsPower tools
Agent Content

Workflow Requirements (MANDATORY)

These requirements apply to ALL geepers agents and orchestrators. Every agent MUST follow these rules.

1. ALWAYS Use TodoWrite

  • Create a todo list at the START of every non-trivial task
  • Track progress by marking items in_progress and completed
  • Break complex tasks into manageable steps
  • This provides visibility and prevents forgetting steps

2. ALWAYS Call Agents

  • EVERY response should consider which geepers agent applies
  • Complex questions → @conductor (master router)
  • Medium tasks → appropriate orchestrator
  • Specific tasks → individual specialist agent
  • Never work alone on infrastructure, deployment, or code changes

3. ALWAYS Commit Before Major Changes

  • Before any significant edit: git add -A && git commit -m "checkpoint before [change]"
  • This creates restore points and prevents lost work
  • "Major" = anything that changes behavior, adds features, or modifies multiple files

4. ALWAYS Read Before Edit

  • Never edit a file without reading it first in this session
  • Prevents blind edits that break surrounding context
  • The Edit tool will fail anyway - this saves wasted attempts

5. ALWAYS Check Existing State First

  • Before creating anything, verify it doesn't already exist
  • gh repo list before creating repos
  • sm status before adding services
  • Glob before Write for new files
  • Prevents duplicates, conflicts, and overwrites

6. Use EnterPlanMode for Multi-File Changes

  • Any task touching 3+ files deserves a plan
  • Gets user sign-off before heavy implementation
  • Prevents wasted work on wrong approaches

7. Parallel Tool Calls When Independent

  • Batch independent reads, searches, and agent launches in single messages
  • Example: Read 5 files simultaneously, not sequentially
  • Dramatically speeds up exploration and reduces round-trips

8. Verify After Changes

  • Run tests after code changes: pytest
  • Check services after changes: sm status
  • Validate Caddy after edits: sudo caddy validate --config /etc/caddy/Caddyfile
  • Screenshot/snapshot for web UI changes

9. Check Recommendations Before Starting

  • Look at ~/geepers/recommendations/by-project/ for existing analysis
  • Don't reinvent work that's already been done
  • Build on previous insights rather than starting fresh

10. Use Explore Agent for Open-Ended Search

  • Direct Grep/Glob for needle queries ("find class Foo", "where is config.py")
  • Task with subagent_type=Explore for discovery ("how does auth work?", "what's the architecture?")
  • Agents handle multi-step exploration better than manual iteration

Quick Reference Checklist

Before starting any task:

  • Check ~/geepers/recommendations/by-project/ for existing analysis
  • Create TodoWrite list if task has multiple steps
  • Consider which agent(s) to involve

Before creating files/services/ports:

  • Glob/search to verify it doesn't exist
  • Check sm status for service conflicts
  • Check Caddyfile for port conflicts

Before editing files:

  • Read the file first
  • Commit current state if making major changes

After making changes:

  • Run relevant tests
  • Verify services are running
  • Validate configs

These rules are also in ~/.claude/CLAUDE.md and are enforced by conductor.

Similar Agents
conversation-analyzer
2 tools

Use this agent when analyzing conversation transcripts to find behaviors worth preventing with hooks. Examples: <example>Context: User is running /hookify command without arguments user: "/hookify" assistant: "I'll analyze the conversation to find behaviors you want to prevent" <commentary>The /hookify command without arguments triggers conversation analysis to find unwanted behaviors.</commentary></example><example>Context: User wants to create hooks from recent frustrations user: "Can you look back at this conversation and help me create hooks for the mistakes you made?" assistant: "I'll use the conversation-analyzer agent to identify the issues and suggest hooks." <commentary>User explicitly asks to analyze conversation for mistakes that should be prevented.</commentary></example>

80.6k
Stats
Stars1
Forks1
Last CommitMar 16, 2026