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/mrinalwadhwa/fluent/agents-mdgit clone --depth 1 https://github.com/mrinalwadhwa/fluentWhat 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 | $0.01247 | $0.01247 |
| Opus 5 | $0.00624 | $0.00624 |
| Sonnet 5 | $0.00249 | $0.00249 |
| Haiku 4.5 | $0.00125 | $0.00125 |
Grade A, and why
fluent AGENTS.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 — 157 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Instructions for Coding Agents
These instructions define how Coding Agents should assist with this project.
Fluent workflow
Use the fluent to build the fluent. For non-trivial code,
documentation, skill, expertise, or behavior changes, follow the
fluent skill and go through the Fluent lifecycle:
brief, behaviors, approach, plan, execution, review, and land.
Do not implement substantial product/code changes directly on main.
Use Fluent Work Items and Attempts for delegated build work that needs
isolation, reviewers, and merging. The Work model path is: Work Item,
Attempt, Task, Workspace, and Merge Candidate.
Conversation agents may edit Fluent planning and memory state directly when they are collaborating with the user in the discussion loop: observations, briefs, behavior drafts, approaches, plans, lightweight curation, and similar durable notes. These edits are part of shaping work, not delegated execution.
Do not meddle with live execution state directly: branches, worktrees, statuses, session artifacts, and merging state belong to the execution system. Modify them only during explicit recovery with the user.
Keep main available as a stable integration branch for Work to rebase
from and merge into. If conversation-state edits could overlap with
active Work or merging, make them on a lightweight discussion branch
or worktree and land them separately instead of dirtying main.
Use fluent observations add to record future work and lessons.
Open observations live as one file per entry under
.fluent/observations/; resolved ones move to
.fluent/observations/resolved/. Inspect with fluent observations list and fluent observations show <id>.
Commit messages
Subject line
- Use imperative mood and active voice
- Start with a verb: "Add", "Fix", "Update", "Remove", "Refactor"
- Keep under 50 characters
- Capitalize the first letter
- Do not end with a period
- Describe the change, not the process that led to it
- Good: "Fix sandbox worktree binding"
- Bad: "Run review and fix issues found"
- Bad: "Address reviewer findings"
- Use "Improve" over "Fix" when the change enhances something that was working but could be better. "Fix" implies it was broken.
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 · 157 lines · 1,247 tokens per session scan A 67f2ac57e82c
fluent AGENTS.md is an instructions file published in the GitHub repository mrinalwadhwa/fluent (84 stars, last pushed 18d ago), licensed Apache-2.0. It adds 1,247 tokens to every session, about $0.0062 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
jarvis CLAUDE.md
Instructions for pguilp25/jarvis, covering claude.md, what jarvis is, commands, run the agent and records carry kind=draft|final + the plan text}.
designing-real-world-ai-agents-workshop CLAUDE.md
Instructions for iusztinpaul/designing-real-world-ai-agents-workshop, covering project, project structure, tech stack, access documentation and running qa.
agent-claim-network AGENTS.md
Instructions for FTShare-Lab/agent-claim-network, covering agents.md, 项目概览, 当前架构, rust 规范 and 数据与安全边界.
SeekMoney-ai CLAUDE.md
Instructions for zykooooooooo/SeekMoney-ai, covering claude.md, project overview, development commands, development and production.
agent-eve CLAUDE.md
Instructions for shehryarsaroya/agent-eve, covering claude.md — agent eve (né the compact, then agent transfer), 0. 🚨 hard rules, 1. what this is, 2. doc map and 3. state of play (2026-07-26).
p2pclaw-mcp-server CLAUDE.md
Claude Code instructions for Agnuxo1/p2pclaw-mcp-server, covering railway api (p2pclaw-mcp-server), critical: string literal newlines in index.js, railway branches, groq api keys (logprobs restriction) and vercel frontend (openclaw-p2p-launch / p2pclaw-v3).