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/beeltec/context-usage-mcp/agents-mdgit clone --depth 1 https://github.com/beeltec/context-usage-mcpWhat 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 | $0.00580 | $0.00580 |
| Opus 5 | $0.00290 | $0.00290 |
| Sonnet 5 | $0.00116 | $0.00116 |
| Haiku 4.5 | $0.00058 | $0.00058 |
Grade A, and why
context-usage-mcp 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 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 — 51 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Guidelines
Project Wiki
This project keeps durable, cross-session knowledge in a wiki at docs/wiki/,
written in the Open Knowledge Format.
Start at docs/wiki/index.md — it links every page.
When to consult it
- At the start of any non-trivial task. Read
docs/wiki/index.md, then the pages relevant to what you're about to change.overview.mdholds the goal, architecture, and every locked decision;log.mdis the chronological decision history. - Before making a design or architecture decision, check whether it was
already resolved. Many choices (transcript-reader vs sidecar, npm publishing,
Codex host adapter, output shape) are already locked in
overview.md— do not silently re-decide them. - When something in the code surprises you. The wiki usually records the why behind non-obvious choices (e.g. why cached tokens are subtracted, why tool schemas use zod raw shapes).
When to update it
Update the wiki when a change produces knowledge that outlives the current task — not for routine code changes the repo and git history already capture. Update after:
- Making or reversing a design/architecture decision → record it in
overview.mdand add a dated entry tolog.md. - Learning a durable, non-obvious fact about a host, dependency, or external system (Claude Code / Codex internals, the MCP SDK, npm publishing) → put it on the relevant page.
- Adding or renaming a wiki page → link it from
index.md.
Do not wiki-fy: transient task state (that belongs in docs/tasks/),
anything derivable from the code, or per-conversation details.
How to update it
- Invoke the
wikiskill (/wiki) for any create/update/organize work so the Open Knowledge Format conventions are applied consistently. - Preserve existing knowledge and edit the smallest relevant set of pages. Prefer updating an existing page over creating a new one.
- Keep OKF front matter intact (
type,title,description,timestamp); match the style of neighbouring pages. - Use root-relative links between pages (e.g.
/docs/wiki/overview.md). - Log entries are append-only: add a new
## YYYY-MM-DD — <summary>section tolog.mdrather than rewriting past history; cross-link to the page that holds the detail.
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 · 51 lines · 580 tokens per session scan A 1cd8edf294ee
context-usage-mcp AGENTS.md is an instructions file published in the GitHub repository beeltec/context-usage-mcp (0 stars, last pushed 1mo ago), licensed MIT. It adds 580 tokens to every session, about $0.0029 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 2026-08-31.
Other instructions, from other repositories
datalens-dev-mcp AGENTS.md
Instructions for ADIKANT/datalens-dev-mcp, covering agents.md, scope, read order, local material policy and public change hygiene.
printyourduck-mcp AGENTS.md
AGENTS.md instructions for printyourduck/printyourduck-mcp, covering agent guide, product boundary, commands, repo ownership and release truth.
homelab-asset-db-mcp CLAUDE.md
Claude Code instructions for McCavity/homelab-asset-db-mcp, covering homelab-asset-db-mcp — conventions for ai agents, what this project is, key design choices, parser contract and coding conventions.
wikipedia-mcp-server AGENTS.md
AGENTS.md instructions for cyanheads/wikipedia-mcp-server, covering developer protocol, what's next?, core rules, patterns and tool.
wikipedia-mcp-server CLAUDE.md
Claude Code instructions for cyanheads/wikipedia-mcp-server, covering developer protocol, what's next?, core rules, patterns and tool.
llm-context.py CLAUDE.md
Instructions for cyberchitta/llm-context.py, covering claude.md, working notes (gitignored) and draining the field notes.