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/foxglove/mcap/agents-mdgit clone --depth 1 https://github.com/foxglove/mcapWhat 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 | $0.01480 | $0.01480 |
| Opus 5 | $0.00740 | $0.00740 |
| Sonnet 5 | $0.00296 | $0.00296 |
| Haiku 4.5 | $0.00148 | $0.00148 |
Grade A, and why
mcap 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 — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Development Guide
Overview
This is a polyglot library monorepo for the MCAP log file format, with implementations in TypeScript, Python, Go, Rust, C++, and Swift. There are no running services — this is purely a library/SDK project. The core functionality is writing and reading MCAP files.
MCAP is a modular container file format for recording timestamped pub/sub messages with arbitrary serialization formats. It is designed to work well under various workloads, resource constraints, and durability requirements. The format specification lives in website/docs/spec/index.md, with the well-known registry in website/docs/spec/registry.md, and feature notes in website/docs/spec/notes.md.
Design principles
Bounded memory when reading. Neither the language libraries nor the CLI may read (or force a consumer to read) an entire MCAP file into memory — files can be many GB. Reader memory should scale with the record or chunk being processed, not with the file length: holding one record, chunk, or attachment at a time is fine, but buffering the whole file, or all of its messages, is an out-of-memory foot-gun.
- Memory-map (
mmap) seekable local files where the language supports it, or have the API consumer supply the bytes (e.g. via their own mmap); use seek + bounded range reads for random access and streaming for sequential scans. - When input isn't seekable (e.g. a stdin pipe) or an operation needs random access over a stream (e.g. sorting), spool to a temporary file and mmap it rather than buffering in memory. A tmpfs temp dir keeps the spool resident, though that is typically swap-backed.
General prerequisites
- Git LFS — test data under
tests/conformance/data/andrust/mcap/tests/data/is stored in Git LFS. Tests will fail withInvalidMagicerrors if LFS pointers haven't been pulled. Rungit lfs pullbefore running tests. - Conformance tests orchestrate cross-language testing via the TypeScript harness:
yarn test:conformance:generate-inputs && yarn test:conformance. They require Git LFS data and pre-built binaries for each target language. - Releasing — see RELEASING.md for the release process for each language.
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 · 109 lines · 1,480 tokens per session scan A a74122b5705f
mcap AGENTS.md is an instructions file published in the GitHub repository foxglove/mcap (1,064 stars, last pushed 6d ago), licensed MIT. It adds 1,480 tokens to every session, about $0.0074 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
text-to-cad AGENTS.md
AGENTS.md instructions for earthtojake/text-to-cad, covering agents.md, branch and layout first, release workflow, repo map and repo rules.
spiceai copilot-instructions.md
Copilot instructions for spiceai/spiceai, covering spice.ai agent instructions, data correctness — absolute top priority, evidence — no claim without a reproduction, build, test, lint (expensive — read first) and git & prs.
quant-mind AGENTS.md
Instructions for LLMQuant/quant-mind, covering quantmind — agent instructions, progressive context loading, what this is, positioning and module map.
riko CLAUDE.md
Claude Code instructions for nerevu/riko, covering riko, key paths, docs map, async backend and correctness invariants (audit remediation).
spiceai AGENTS.md
AGENTS.md instructions for spiceai/spiceai, a project described as: Add a real-time analytics node to your operational database. Spice is a portable, accelerated SQL query, search, and LLM-inference engine in Rust for data-grounded AI apps and agents.
autonomous-os AGENTS.md
Instructions for autonomous-ai/autonomous-os, covering agents.md, multi-ide rules, working style, parallel work / subagents and device access rules.