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/andyzengmath/quantum-loop/claude-mdgit clone --depth 1 https://github.com/andyzengmath/quantum-loopWhat 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 | $0.06660 | $0.06660 |
| Opus 5 | $0.03330 | $0.03330 |
| Sonnet 5 | $0.01332 | $0.01332 |
| Haiku 4.5 | $0.00666 | $0.00666 |
Grade A, and why
quantum-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 2d 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 — 499 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Quantum-Loop: Agent Instructions
You are an autonomous implementation agent in the quantum-loop system. Each invocation gives you a fresh context with no memory of previous iterations. All state is in quantum.json.
Step 1: Read State
- Read
quantum.jsonin the current directory - Read the
codebasePatternsarray for project conventions discovered by previous iterations - Read the PRD at the path specified in
prdPathfor requirement context - Check the
progressarray for recent learnings
Parallel Mode (Worktree Execution)
Check if your working directory is inside .ql-wt/. If so:
- Do NOT write quantum.json -- the orchestrator manages all state. Only the orchestrator reads and writes quantum.json.
- You MUST commit your changes before signaling completion:
git add -A && git commit -m "feat: <Story ID> - <Story Title>". The orchestrator merges committed branches — uncommitted work is lost when the worktree is removed. - Signal completion via stdout only using
<quantum>STORY_PASSED</quantum>or<quantum>STORY_FAILED</quantum>. - Your story ID is provided in the prompt argument, not inferred from quantum.json. Implement only the story you were assigned.
If you are NOT in a worktree (i.e., running in the repo root), follow the standard sequential process below.
Step 2: Verify Branch
The correct branch is specified in quantum.json.branchName.
git branch --show-current
If you're not on the correct branch:
git checkout <branchName> 2>/dev/null || git checkout -b <branchName> main
Step 3: Select Story
Find the next executable story from the dependency DAG:
Eligible stories must satisfy ALL of:
statusis"pending"OR"failed"(withretries.attempts < retries.maxAttempts)- ALL stories in
dependsOnhavestatus: "passed"
Among eligible stories: pick the one with the lowest priority number.
Defensive check: If you encounter a story with status: "in_progress" that was NOT assigned to you (i.e., you are not the agent responsible for it), do NOT modify it. Log: "WARNING: Story <ID> is in_progress but not assigned to this agent — skipping." If you are in worktree mode and your assigned story is in_progress with startedAt older than 10 minutes, this is likely a retry after a stale detection — proceed normally with implementation.
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.
- 2d ago First seen · 499 lines · 6,660 tokens per session scan A fcc3446996b5
quantum-loop CLAUDE.md is an instructions file published in the GitHub repository andyzengmath/quantum-loop (24 stars, last pushed 2mo ago), licensed MIT. It adds 6,660 tokens to every session, about $0.0333 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 instructions, from other repositories
vast-cli CLAUDE.md
Instructions for vast-ai/vast-cli, covering vast-ai/vast-cli and ai agents.
claude-ads CLAUDE.md
Instructions for AgriciDaniel/claude-ads, covering claude ads repository instructions, architecture, development and verification.
notebooklm-skill AGENTS.md
Instructions for claude-world/notebooklm-skill, covering notebooklm-skill, authentication, cli commands, notebooklm-skill — core operations and notebooklm-pipeline — workflow orchestration.
stockbit-mcp CLAUDE.md
Instructions for INo-xious/stockbit-mcp, covering claude.md, what this is, commands, the map and three invariants. do not break them; each has a test.
devin-handoff AGENTS.md
Instructions for club-cog/devin-handoff, covering devin handoff — agent guide, what this is, when to hand off, finding the script and how to use.
agent-plugins-skills copilot-instructions.md
Instructions for richfrem/agent-plugins-skills, covering copilot instructions for agent-plugins-skills, 1. think before coding, 2. simplicity first, 3. surgical changes and 4. goal-driven execution.