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/chroma-core/chroma/claude-mdgit clone --depth 1 https://github.com/chroma-core/chromaWhat 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 | $0.00535 | $0.00535 |
| Opus 5 | $0.00267 | $0.00267 |
| Sonnet 5 | $0.00107 | $0.00107 |
| Haiku 4.5 | $0.00053 | $0.00053 |
Grade A, and why
chroma 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 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.
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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Chroma Codebase Guidelines
Commit Message Style
Use the [TYPE](scope): Description format:
[ENH](rust-client): Add retry middleware
[BUG](rust-client): Fix connection timeout on retry
[TST](gc): Add MCMR hard delete integration test
[DOC](api): Update embeddings endpoint docs
[CHORE](deps): Bump reqwest to 0.13
Common types: ENH (feature/enhancement), BUG (fix), TST (test), DOC
(docs), CHORE (maintenance/refactor), BLD (build system changes).
Scope is optional but encouraged — use the component name (e.g.
rust-client, gc, dashboard-api).
50/72 Rule
- Subject line: ≤ 50 characters. Keeps it readable in
git log --oneline, GitHub PR lists, and rebase tooling. - Blank line between subject and body (required — many tools use this to split them).
- Body lines: wrap at 72 characters. Leaves room for indentation in 80-column terminals and email patch workflows.
- Imperative mood in the subject: "Add login command", not "Added" or "Adds". Matches Git's own generated messages.
- Body explains what and why, not how — the diff shows how.
With the [TYPE](scope): prefix, aim to keep the whole subject under 50 chars.
If the scope makes that tight, drop it — the type alone is fine.
Tilt-Backed Tests
Tests named test_k8s_integration depend on a local Tilt-managed Kubernetes
environment and should not be run blindly.
Before running these tests, ensure Tilt is already running and the local services are ready. In this repo, the key readiness signals are:
- frontend healthcheck responds on
http://localhost:8000/api/v2/healthcheck - Spanner emulator is reachable on
localhost:9010 - required Kubernetes pods in the
chromanamespace areReady
If a task requires running test_k8s_integration tests and Tilt is not
clearly ready yet, do one of these before running the tests:
- verify an existing Tilt environment is ready and then run the tests
- start or restart Tilt, wait for readiness, and only then run the tests
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 First seen · 56 lines · 535 tokens per session scan A 36fbe3342bf0
chroma CLAUDE.md is an instructions file published in the GitHub repository chroma-core/chroma (29,182 stars, last pushed 3d ago), licensed Apache-2.0. It adds 535 tokens to every session, about $0.0027 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
lmnr CLAUDE.md
Instructions for lmnr-ai/lmnr, covering claude.md, project overview, repository structure, development commands and frontend (next.js).
azure-sdk-for-rust pwsh.instructions.md
Instructions for Azure/azure-sdk-for-rust, covering powershell script instructions, path handling, command invocation, output grouping and console output.
azure-sdk-for-rust resourcemanager.instructions.md
Instructions for Azure/azure-sdk-for-rust, a project described as: This repository is for the active development of the Azure SDK for Rust. For consumers of the SDK we recommend visiting Docs.rs and looking up the docs for any of libraries in the SDK.
azure-sdk-for-rust github-pullrequest.instructions.md
Instructions for Azure/azure-sdk-for-rust, covering instructions for github pull requests and code requirements.
azure-sdk-for-rust git-commit.instructions.md
Instructions for Azure/azure-sdk-for-rust, a project described as: This repository is for the active development of the Azure SDK for Rust. For consumers of the SDK we recommend visiting Docs.rs and looking up the docs for any of libraries in the SDK.
agentgateway copilot-instructions.md
Instructions for agentgateway/agentgateway: Do not check for, speculate about, or report compilation errors during code review. Compilation diagnostics from review are frequently incorrect; rely on CI to detect and report compilation failures.