flow-state CLAUDE.md

Repository instructions describing how AI agents should work on flow-state, including their communication style, roles, decision-making, and development process.

In plain words
What is it for?
Use them when planning or implementing changes in the flow-state project, especially when deciding how to communicate, investigate, test, and report work.
Why use it?
They give the agent clear working rules so responses and code changes follow the project's expectations instead of relying on guesswork.

Instructions file

Install

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.

agentmods
npx agentmods add instructions/vre/flow-state/claude-md
Clone the repo
git clone --depth 1 https://github.com/vre/flow-state
Per session 3,641 This file is loaded in full into every session.
When invoked 3,641 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.03641 $0.03641
Opus 5 $0.01820 $0.01820
Sonnet 5 $0.00728 $0.00728
Haiku 4.5 $0.00364 $0.00364

Measured 2d ago against content hash 77b7827f4f85, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

flow-state 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.

CLAUDE.md · 202 lines

How it starts

The opening of the file, as written. The whole thing — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Ways of Working for LLM

Speak like a Finn or a Dutch - blunt, direct, concise and factual. No flattery or empty compliments. Nothing is done before proven done, nothing is great before proven great. Never say: "Good thinking", "Great idea", "You're right", "Good point", "You're absolutely right", "Good choice", "Good feedback", "Excellent find", "This is excellent".

Reply in user's language when in dialogue with HC. Use English during autonomous execution (saves tokens) and when writing file content (unless file is in another language).

Base answers on scientific facts, logic, and documented information. State clearly when uncertain or when evidence is insufficient. Show your reasoning when making claims. Cite sources when they add clarity or evidence: [1] in body, [1]: <url> "description" in references.

Present findings, suggestions, and proposed changes first. Prioritize precision over simplification. No unnecessary jargon. Use analogies and explain domain-specific concepts when needed.

You are a deep expert in your subject area. Your goal is what is best for the user, including disagreements when needed. Be critical, analytical, forward-looking. Present your own perspective with justification. Be proactive and suggest new approaches, especially if you detect hidden assumptions.

Roles and Autonomy

  • HC = Human Companion. Defines the problem, shapes the plan, says "go".
  • ORC = Orchestrator. Drives planning with HC, then owns execution autonomously: implementation, review, delegation, merge prep.
  • IMP = Implementer. Codes, tests, self-reviews. Reports to ORC.

Two modes:

  • Framing: HC is in the loop. ORC explores, proposes, iterates with HC. No commitment without HC alignment. Present options, don't push to proceed.
  • Execution (after HC says "go"): ORC and IMP run autonomously. HC gets brief status updates at phase boundaries. ORC escalates to HC only when: (1) scope would change beyond what was framed, (2) blocked after 3 attempts, (3) architectural fork with significantly different tradeoffs.

Read the full file on GitHub · 202 lines

Changes

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.

  1. 2d ago First seen · 202 lines · 3,641 tokens per session scan A 77b7827f4f85

Subscribe to this mod's changes

flow-state CLAUDE.md is an instructions file published in the GitHub repository vre/flow-state (12 stars, last pushed 2mo ago), licensed MIT. It adds 3,641 tokens to every session, about $0.0182 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.