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/curie-eng/curie/agents-mdgit clone --depth 1 https://github.com/curie-eng/curieWrote 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/curie-eng/curie/agents-md)<a href="https://agentmods.dev/instructions/curie-eng/curie/agents-md"><img src="https://agentmods.dev/badge/instructions/curie-eng/curie/agents-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 5.1 | $0.10866 | $0.10866 |
| Opus 5 | $0.05433 | $0.05433 |
| Sonnet 5 | $0.02173 | $0.02173 |
| Haiku 4.5 | $0.01087 | $0.01087 |
Grade C, and why
curie AGENTS.md scanned grade C with 2 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.
Hidden instructionshighPrompt injection
Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.
`<!-- doclint:ignore-line -->` on that line (or the line before it). Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
if curl -fsS http://localhost:23000/api/public/health >/dev/null 2>&1; then How it starts
The opening of the file, as written. The whole thing — 737 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md - Curie
Agent instructions for this repo. Start with README.md for what
the product is and how to run it, and ARCHITECTURE.md for
the component diagram, the message-flow sequence (Slack mention -> dispatcher ->
worker -> sandbox -> runner -> Slack reply), and the deploy-flow sequence (git
push -> webhook -> bundle pipeline -> deployment). The one-line version: a Slack
message is answered by a versioned plugin running in an isolated Kubernetes
sandbox, traced through Langfuse and steerable mid-turn; a git push deploys that
plugin under a bot identity via the API's git-flow engine. Relay is the project
codename; curie is the product-surface name (bot handle, CLI binary). Read
ARCHITECTURE.md before touching a cross-component seam. If you are a coding
agent orienting in this repo, llms.txt is the curated machine map
of these docs, organized around the parity ladder. This file is for an agent
working on this repository; an agent driving Curie (a released binary,
someone else's bundle) wants docs/agents.md, the
verification contract that names the exact command proving each outcome. The two questions this repo
exists to answer are
why did my agent work locally but break once deployed?
and
how do I test an agent the same way locally and on Kubernetes? —
the same immutable bundle and the same evals/cases.json across skill,
local, and cluster is the answer to both.
Directory map
One directory is one ownership boundary. Each area's own CLAUDE.md (linked
below) carries the rules and verify commands specific to that area -- read it
before editing there, in addition to this file.
| Path | Language | Scoped rules |
|---|---|---|
packages/aci-protocol |
Python (Pydantic + codegen) | packages/CLAUDE.md |
packages/plugin-format |
Python (Pydantic + codegen) | packages/CLAUDE.md |
apps/api |
Python (FastAPI) | apps/api/CLAUDE.md |
apps/dispatcher |
Python (Slack Bolt) | apps/dispatcher/CLAUDE.md |
apps/mail-adapter |
Python (stdlib HTTP + Pydantic) | apps/mail-adapter/CLAUDE.md |
apps/worker |
Python (redis-py) | apps/worker/CLAUDE.md |
runner |
Python (claude-agent-sdk) | runner/CLAUDE.md |
apps/ui |
React (Vite + TS) | apps/ui/CLAUDE.md |
cli |
Rust (clap + tokio) | cli/CLAUDE.md |
charts/curie |
Helm | charts/curie/CLAUDE.md |
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 Changed · +21 lines · +414 tokens per session 28c78b85c345
- 2d ago Changed · +35 lines · +548 tokens per session 158b7356bfd3
- 6d ago First seen · 681 lines · 9,904 tokens per session scan C aa690d65c0ec
curie AGENTS.md is an instructions file published in the GitHub repository curie-eng/curie (35 stars, last pushed yesterday), licensed Apache-2.0. It adds 10,866 tokens to every session, about $0.0543 per session on Opus 5. A static security scan graded it C with 2 findings (hidden instructions, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
platform-skills copilot-instructions.md
Copilot instructions for nitinjain999/platform-skills, covering platform engineering instructions for github copilot, version: 1.40.0, source: https://github.com/nitinjain999/platform-skills, response contract and core principles.
recursive-decomposition-skill AGENTS.md
Instructions for massimodeluisa/recursive-decomposition-skill, covering agents: recursive-decomposition skill, read these first (mandatory), language policy, non-negotiables and commands.
stockbit-mcp CLAUDE.md
Claude Code 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.
embedded-systems-architect AGENTS.md
Instructions for Zachariah9420/embedded-systems-architect: This repository ships an agent skill for embedded-systems and edge-AI development work.
agentic-validation-skills CLAUDE.md
Claude Code instructions for vivekkrishna/agentic-validation-skills, covering agentic test automation guidelines (cige), 1. establish context before testing, 2. anchor every test to intent, 3. define guardrails before executing and 4. treat execution as adaptive guidance.
octowiz CLAUDE.md
Instructions for raelli/octowiz, covering octowiz, releasing & deploying, commands, agent skills and issue tracker.