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 skills add joaquimscosta/arkhe-claude-plugins --skill ralph-loopgit clone --depth 1 https://github.com/joaquimscosta/arkhe-claude-pluginsWrote 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/joaquimscosta/arkhe-claude-plugins/ralph-loop)<a href="https://agentmods.dev/skills/joaquimscosta/arkhe-claude-plugins/ralph-loop"><img src="https://agentmods.dev/badge/skills/joaquimscosta/arkhe-claude-plugins/ralph-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/joaquimscosta/arkhe-claude-plugins/ralph-loop"><img src="https://agentmods.dev/badge/skills/joaquimscosta/arkhe-claude-plugins/ralph-loop.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00087 | $0.00761 |
| Opus 5.5 | $0.00035 | $0.00304 |
| Sonnet 5.5 | $0.00017 | $0.00152 |
| Haiku 4.5 | $0.00009 | $0.00076 |
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 yesterday.
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 — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ralph Loop Execution Protocol
Autonomous development with fresh context per iteration and Hat-lite builder/verifier workflow.
Quick Start
# From project directory (after /create-prd)
./ralph.sh 20 # Run up to 20 iterations
./ralph.sh 5 # Quick test with 5 iterations
How It Works
Each iteration runs in a fresh context window:
┌─────────────────────────────────────────────────────┐
│ ITERATION (Fresh Context) │
├─────────────────────────────────────────────────────┤
│ 1. ORIENT: Read activity.log + tasks.json + memories│
│ 2. BUILD: Pick ONE task, implement, verify │
│ 3. VERIFY: Review, update status, commit │
│ 4. LEARN: (Optional) Save insights to memories │
│ 5. DECIDE: All done? → RALPH_COMPLETE │
└─────────────────────────────────────────────────────┘
Hat-Lite System
| Role | When | Responsibility |
|---|---|---|
| Builder | ~65% of iteration | Implement ONE task |
| Verifier | ~35% of iteration | Review, update, commit |
Key Files
| File | Purpose |
|---|---|
.ralph/current-taskset/tasks.json |
Task state (source of truth) |
.ralph/current-taskset/activity.log |
Iteration history |
.ralph/current-taskset/memories.md |
Persistent learnings |
.ralph/current-taskset |
Symlink to active task set |
PROMPT.md |
Instructions per iteration |
ralph.sh |
Loop runner script |
Task JSON Format
Tasks are stored in .ralph/current-taskset/tasks.json:
{
"tasks": [
{
"id": "setup-001",
"description": "Initialize project",
"verificationTier": "build",
"passes": false,
"iteration_completed": null
}
]
}
Valid verificationTier values: "build" (default), "visual", "api", "e2e". See WORKFLOW.md for details.
Completion Signal
When all tasks pass, output:
RALPH_COMPLETE: All tasks verified
Commands
What ships with it
4 files 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.
- yesterday First seen · 95 lines · 87 tokens per session scan A 5e01ebf25d9f
ralph-loop is a skill published in the GitHub repository joaquimscosta/arkhe-claude-plugins (21 stars, last pushed yesterday), licensed MIT. It adds 87 tokens to every session and 761 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-10-16.
Other skills, from other repositories
git-workflow
Git branching, commits, and PR workflow following GitFlow conventions.
dev-finish
Close out a session — summarise what changed, capture what was learned, and prepare the commit.
branch-workflow
Branch and commit conventions that keep an agent's work reviewable — one branch per task, main stays clean, nothing force-pushed.
claim
Acquire, renew or release a Fakoli State task lease when beginning work or recovering an interrupted claim.
undercover-mode
When contributing to public or open-source repositories where AI attribution should be hidden. Use when the user says "stealth mode," "undercover," "no AI attribution," "hide AI," "contribute anonymously," "clean commits," "no co-authored-by," or when working on repos where AI involvement should not be visible. Strips…
java-commit
Generates a Conventional Commits message for staged Java changes. Use when user asks to "write a commit message", "help me commit", "what should my commit say", "summarize my changes", "draft a commit", or "create commit message".