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 instructions/rmolines/fractal-loop/claude-mdgit clone --depth 1 https://github.com/rmolines/fractal-loopWrote 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/instructions/rmolines/fractal-loop/claude-md)<a href="https://agentmods.dev/instructions/rmolines/fractal-loop/claude-md"><img src="https://agentmods.dev/badge/instructions/rmolines/fractal-loop/claude-md.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 | $0.01486 | $0.01486 |
| Opus 5 | $0.00743 | $0.00743 |
| Sonnet 5 | $0.00297 | $0.00297 |
| Haiku 4.5 | $0.00149 | $0.00149 |
Grade A, and why
fractal-loop CLAUDE.md 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 3d 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 — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What this is
Fractal Loop — recursive project management for Claude Code. A plugin that decomposes goals into predicates and works on the riskiest unknown first. One operation, repeated at any scale.
Version: 0.9.0 | Plugin manifest: .claude-plugin/plugin.json
Commands
# Test the state machine
bash scripts/fractal-state.sh
# View tree state (auto-discovers tree in .fractal/)
bash scripts/fractal-tree.sh
# Read active predicate
bash scripts/active-predicate.sh
# Timeline: concluded nodes newest-first
bash scripts/fractal-timeline.sh
No build step. No dependencies. Pure shell scripts + markdown skills.
Architecture
The primitive
fractal(predicate):
evaluate(predicate, existing_children) → new_child | complete | leaf | unachievable
if unachievable → prune
if leaf → specify → execute → human validates
if new_child → create child → human validates → recurse → re-evaluate parent
if complete → validate branch satisfaction or select next pending child
Skill chain (execution order)
/fractal:init— bootstrap: extract objective, create tree/fractal:run— idempotent state machine (main entry point, call repeatedly)/fractal:patch— fast path for trivial leaf predicates- Sprint agent (
agents/sprint.md) — runs the full cycle/fractal:planning→/fractal:delivery→/fractal:review→/fractal:shipas a single Sonnet subagent with no human gates /fractal:doctor— tree integrity validation/standards:generate— inspect repo and generate.claude/standards.md(called by/fractal:initwhen no standards.md exists)
Skills live in commands/. Subagents live in agents/ (evaluate, sprint, patch-worker).
On-disk state (.fractal/ in target repo)
The filesystem IS the state. No database, no JSON.
root.md— root predicate +active_nodepointer (always exactly one per tree)predicate.md— per node: verifiable condition, status (pending|satisfied|pruned)discovery.md— per node (after evaluation): evaluator response (new_child|complete|leaf|unachievable). Ephemeral on parents — deleted when child ascends.prd.md— leaf nodes only: acceptance criteria, out-of-scope, constraints- Execution state derived from artifact presence:
- Only
predicate.md→ not started discovery.mdexists (new_child) → create proposed child, recursediscovery.mdexists (leaf, no prd) → specify, write prddiscovery.md+prd.md→ specified (run sprint)plan.mdexists → planned (run delivery)plan.md+results.md→ executed (run review)plan.md+results.md+review.md→ reviewed (validate, then ship)
- Only
conclusion.md— per node (after satisfaction): what was achieved, key decisions, deferred itemslearnings.md— accumulated human corrections (read on every/fractal:runentry)
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.
- 3d ago First seen · 117 lines · 1,486 tokens per session scan A 0d499c1db5b0
fractal-loop CLAUDE.md is an instructions file published in the GitHub repository rmolines/fractal-loop (13 stars, last pushed 5mo ago), licensed MIT. It adds 1,486 tokens to every session, about $0.0074 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 2026-08-31.
Other instructions, from other repositories
project-orchestration-skills CLAUDE.md
Instructions for jrjsmrtn/project-orchestration-skills, covering ai-assisted project orchestration skills, project structure, skill file format, skills overview and development conventions.
STAR AGENTS.md
AGENTS.md instructions for wanghao9610/STAR, covering agent instructions, 1. think before coding, 2. simplicity first, 3. surgical changes and 4. goal-driven execution.
no-vibe AGENTS.md
Instructions for rizukirr/no-vibe, covering agents, repo purpose, architecture (high-signal files), verification commands and conventions that are easy to miss.
pamplejuce AGENTS.md
Instructions for sudara/pamplejuce, a project described as: A JUCE audio plugin template. JUCE 8, Catch2, Pluginval, macOS notarization, Azure Trusted Signing, Github Actions.
openclaw-mcp-bridge CLAUDE.md
Instructions for gabrielekarra/openclaw-mcp-bridge, covering claude.md, project overview, build & development commands, architecture and key dependencies.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).