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/tabnas/mcp/agents-mdgit clone --depth 1 https://github.com/tabnas/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 | $0.04275 | $0.04275 |
| Opus 5 | $0.02138 | $0.02138 |
| Sonnet 5 | $0.00855 | $0.00855 |
| Haiku 4.5 | $0.00428 | $0.00428 |
Grade C, and why
mcp AGENTS.md scanned grade C with 1 finding 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
rm -rf node_modules/@tabnas/parser node_modules/@tabnas/support How it starts
The opening of the file, as written. The whole thing — 284 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agents Guide — mcp
What this project is
The tabnas agent tooling, published as @tabnas/mcp: an MCP server
(stdio) and the unified tabnas CLI, built as two thin front-ends
over one core so they cannot disagree (admin ADR-10/11, Workstream C).
Six operations only — parse, validate_grammar,
explain_parse_error, test_grammar, list_plugins,
describe_plugin — each a pure plain-JSON-in, plain-JSON-out function.
This repo is TypeScript-only. It is tooling over the engine, not a
parity package: there is no Go port, no test/spec/ fixture contract of
its own, and the engine's no-dependency rule does not apply here
(@modelcontextprotocol/sdk and ajv are regular dependencies —
structural grammar validation is runtime behaviour of
validate_grammar, so Ajv is deliberately NOT a devDependency).
Repository map
| Path | What it is |
|---|---|
ts/src/core.ts |
The six data operations. The ONLY place operation logic lives. |
ts/src/compat.ts |
The seventh operation, grammar compatibility (plan Phase 5). Its own module because it is the only one that loads TWO grammars and runs them against each other; it reuses core's firewall, bounds and instance builder rather than a copy. |
ts/src/mcp.ts |
MCP front-end: the surface from tools.ts, served over stdio only. Package main; importing it must never touch stdio (the transport starts only under require.main). Exports main() — the CLI's mcp subcommand starts the identical server through it. |
ts/src/cli.ts |
The tabnas CLI front-end: argument plumbing, human rendering, exit codes. The mcp subcommand lazily requires mcp.ts and runs the stdio server (so the data-command fast paths never load the MCP SDK). |
ts/src/tools.ts |
The tool + resource surface: TOOLS, RESOURCES, callTool. No transport. Both front-ends import it, so neither drags in the other's transport — the reason the Worker can exist at all. |
ts/src/data.ts |
Accessors for the bundled data. Reads a static import, never the filesystem (the Worker has none). |
ts/src/data-bundle.ts |
Generated, gitignored. data/ compiled into a module by tools/embed-data.js. |
ts/src/grammar-validator.js |
Generated, gitignored. data/grammar.schema.json precompiled by Ajv (tools/build-validator.js), because Workers forbid new Function. |
ts/tools/gen-data.js |
Regenerates data/ from sibling checkouts (../<repo>). npm run gen-data. |
ts/tools/embed-data.js |
Build step: compiles data/ + the package version into ts/src/data-bundle.ts. |
ts/tools/build-validator.js |
Build step: precompiles the grammar schema into ts/src/grammar-validator.js. |
data/ |
Bundled, generated, committed copies of the fleet contract files: grammar.schema.json, diagnostic.schema.json, error-codes.json, DIVERGENCE.md, plugins.json. Never edit by hand. |
ts/test/ |
node --test suites, CJS. golden.test.js is the front-end parity gate. |
benchmark/ |
The AX benchmark (plan E1): ten agent tasks, their starting state, and a machine check per task. --self-test runs as part of npm test and measures the benchmark, not any agent. See benchmark/README.md. |
ts/src/worker.ts |
The hosted endpoint (plan Phase 4): streamable-HTTP MCP at POST /mcp, plus /health and /.well-known/mcp. Transport ONLY — every parsing decision is the same core, so hosted and local cannot diverge. Its exports must all be functions (see below). |
ts/src/budget.ts |
The hosted endpoint's limits and shape-only telemetry. Separate from worker.ts because workerd rejects a non-function named export on a Worker entrypoint. |
wrangler.json |
The hosted Worker's deploy config (mcp.tabnas.dev). Separate from the website's Worker on purpose. |
ci/ci.yml |
The staged CI workflow (see "CI"). |
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 · 284 lines · 4,275 tokens per session scan C 94f11d0f65d5
mcp AGENTS.md is an instructions file published in the GitHub repository tabnas/mcp (0 stars, last pushed 4d ago), licensed MIT. It adds 4,275 tokens to every session, about $0.0214 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.