Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/changoo89/claude-pilot/ralph-loopnpx skills add changoo89/claude-pilot --skill ralph-loopgit clone --depth 1 https://github.com/changoo89/claude-pilotWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/changoo89/claude-pilot/ralph-loop)<a href="https://agentmods.dev/skills/changoo89/claude-pilot/ralph-loop"><img src="https://agentmods.dev/badge/skills/changoo89/claude-pilot/ralph-loop.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00023 | $0.01736 |
| Opus 5 | $0.00012 | $0.00868 |
| Sonnet 5 | $0.00005 | $0.00347 |
| Haiku 4.5 | $0.00002 | $0.00174 |
Grade A, and why
ralph-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 206 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SKILL: Ralph Loop
Purpose: Autonomous completion loop - iterate until all tests pass, coverage met, type-check clean Target: Coder, Tester, Validator teammates ⚠️ Teammates Only: This skill is for coder/tester teammates in Agent Teams. Team Lead spawns teammates who execute this loop autonomously.
Quick Start
When to Use This Skill
- After first code change (Entry point)
- Tests failing or coverage low
- Type-check errors present
- Lint violations found
Quick Reference
# Ralph Loop: Autonomous iteration
iteration=0
max_iterations=7
early_escalation=false
# Check for --early flag or architecture-related failure
if [[ "$*" == *"--early"* ]] || is_architecture_failure; then
max_iterations=2
early_escalation=true
fi
while [ $iteration -lt $max_iterations ]; do
echo "Iteration $((iteration + 1))/$max_iterations"
# Run verification
if run_all_checks; then
echo "✓ All quality gates passed"
break
fi
# Fix failures
fix_failures
# Update state
update_state "$SC" "in_progress" $((iteration + 1))
((iteration++))
done
# Complete or escalate
if [ $iteration -eq $max_iterations ]; then
# Message Team Lead for GPT Architect escalation
echo "Max iterations reached. Messaging Team Lead for GPT Architect escalation..."
else
echo "<CODER_COMPLETE>" # All checks pass
fi
Core Concepts
Entry Point
Trigger: Immediately after first code change
Detection:
- Coder teammate makes first edit/write
- Tests run and fail
- Ralph Loop begins automatically
- TeammateIdle hook prevents premature idle on quality check failure
Quality Gates
All must pass before completion:
- Tests:
npm test- All tests pass - Coverage: ≥80% overall, ≥90% core modules
- Type-check:
npm run type-checkortsc --noEmit - Lint:
npm run lint- Zero violations - TODOs: All SC TODOs must be
[x]before<CODER_COMPLETE>- PLAN_PATH from execute-plan prompt OR auto-detect in
.pilot/plan/in_progress/ grep -q "^- \[ \]" "$PLAN_PATH"must return false (no unchecked items)
- PLAN_PATH from execute-plan prompt OR auto-detect in
- TaskCompleted Hook: Automatically verifies gates 1-5 when teammate marks task done
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 206 lines · 23 tokens per session scan A 9ce10a8dc891
ralph-loop is a skill published in the GitHub repository changoo89/claude-pilot (20 stars, last pushed 6mo ago), licensed MIT. It adds 23 tokens to every session and 1,736 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
github-flow-for-claude-on-web
Complete GitHub workflow for Claude Code on the web. ALL GitHub operations MUST use REST API (never gh CLI). Includes branch naming (claude/-sessionId), push retry logic, PR/issue management via API, and complete workflows. Use for all GitHub interactions in Claude Code web environment.
vibing-orchestrate
Run several independent tasks in parallel by creating worker vibing.nvim chats, handing each one a self-contained brief, and aggregating their results. Use when the user asks for parallel work ("do these three in parallel", "split this across separate chats", "並行でやって", "この2つを別々のチャットで").
test-design
Automatically design comprehensive E2E test cases for newly implemented vibing.nvim features. Use immediately after completing feature implementation (Phase 5.4) and before running E2E tests. Generates test scenarios covering Happy paths, Error cases, Edge cases, and Integration points with priority ranking…
vibing-chat-search
Use when the user wants to find a past vibing.nvim conversation by topic — phrases like "前に〜について聞いたことあったっけ", "did we talk about X before", "find that chat where I asked about Y". Searches every chat file under .vibing/chat/ (both User and Assistant content) for a natural-language query, narrows candidates with grep…
remote-screenshot
Take a real screenshot of a running Neovim from inside the Claude Code on the web container, which has no X server. Use when asked to show what a UI change looks like, or to attach an image of the chat buffer, a split layout, an approval prompt or a picker. Requires CLAUDECODEREMOTE=true — on a local machine there is…
vibing-worktree-attach
Attach the current or a new vibing.nvim chat to an already-existing git worktree via natural language — no separate UI. Use when the user wants to switch/attach a chat to an existing worktree ("let's go into the auth-fix worktree", "attach to worktree X", "continue in the feature-y worktree").