OpenSkillIndex← back to search
claude-code

Awesome-omni-skill ralph

持久化执行-验证-修复循环，直到任务完成

install
source · Clone the upstream repo
git clone https://github.com/diegosouzapw/awesome-omni-skill
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/diegosouzapw/awesome-omni-skill "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/tools/ralph-1molchuan" ~/.claude/skills/diegosouzapw-awesome-omni-skill-ralph-cd3819 && rm -rf "$T"
manifest: skills/tools/ralph-1molchuan/SKILL.md
source content

oh-my-ccg Ralph Mode

"The boulder never stops." Persistent loop that keeps working until verification passes.

Activation

Trigger: "ralph", "don't stop"

Loop Structure

Start (iteration 0)
  │
  ├─ Execute: Run the task (executor agent)
  ├─ Verify: Check results (verifier agent)
  │   ├─ Tests pass?
  │   ├─ Build succeeds?
  │   └─ LSP clean?
  │
  ├─ All pass? → COMPLETE ✅
  └─ Any fail? → Fix issues → Next iteration
      └─ Max iterations (10)? → STOP with report

Implementation Steps

  1. Initialize ralph state: iteration=0, maxIterations=10
  2. Execute the current task using executor agent
  3. Run verification: 
    • npm test
      or equivalent
    • tsc --noEmit
    • LSP diagnostics check
  4. If verification passes: mark complete, deactivate ralph
  5. If verification fails:
    • Increment iteration
    • Analyze failures with debugger agent
    • Apply fixes
    • Return to step 2
  6. If max iterations reached: stop, report remaining issues

State

Tracked in 

.oh-my-ccg/state/ralph-state.json
:

  • iteration, maxIterations
  • lastVerification (tests, build, lsp, issues)
  • active flag

Integration

  • Works inside autopilot (auto-impl uses ralph per task)
  • Works standalone for any execution task
  • Stop hook prevents Claude from stopping while ralph is active