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/lathe-cli/lathe/claude-mdgit clone --depth 1 https://github.com/lathe-cli/latheWhat 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 | $0.01909 | $0.01909 |
| Opus 5 | $0.00955 | $0.00955 |
| Sonnet 5 | $0.00382 | $0.00382 |
| Haiku 4.5 | $0.00191 | $0.00191 |
Grade A, and why
lathe 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 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 — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Project intent
Lathe generates agent-friendly Cobra CLIs from Swagger 2.0, OpenAPI 3, google.api.http protobuf APIs, and policy-curated GraphQL schemas. Declared API specs and project configuration are the source of truth. Pinned Git sources are reproducible; local_path sources intentionally follow the referenced working tree.
Product positioning
Lathe is a spec-to-agent-toolchain generator. It turns declared API specs, from pinned Git sources or explicit local working trees, plus repo-local configuration into one inspectable Cobra CLI that both humans and AI agents can run safely.
The core product promise is: agents should not guess command names, flags, auth state, request body shape, HTTP path, or output format. Generated CLIs must expose machine-readable contracts (search, commands --json, commands show, commands schema) and generated Skill guidance so agents can discover, inspect, verify auth, and then execute.
Lathe also owns lathe init for CLI-first application repositories and lathe skill install for its bundled Agent Skill. These distribution surfaces must serve the spec-to-agent workflow; Lathe is not a generic scaffolder, CLI framework, plugin loader, GUI/TUI, API gateway, or hand-written SDK replacement. Product work should strengthen spec fidelity, reproducibility, generated command correctness, runtime catalog inspectability, auth/body/output behavior, application initialization, and generated Skill quality. Challenge features that move product weight into manually authored commands, runtime plugins, or unrelated scaffolding.
Source of truth
- Follow this file for repository structure, commands, style, tests, commits, PR rules, and release workflow entry points.
- Follow
CONTRIBUTING.mdfor contributor workflow and public contribution scope. - Trust code and current Makefile targets over stale documentation when they disagree.
- In this repository, do not commit generated output under
internal/generated/, upstream clones under.cache/, example build artifacts, or ad-hoc generatedskills/<cli-name>/directories. Downstream repositories created bylathe inithave their own generated-output policy.
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 · 99 lines · 1,909 tokens per session scan A 97818165b065
lathe CLAUDE.md is an instructions file published in the GitHub repository lathe-cli/lathe (46 stars, last pushed 2d ago), licensed Apache-2.0. It adds 1,909 tokens to every session, about $0.0095 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
scalar AGENTS.md
Instructions for scalar/scalar, covering agents.md - ai agent guide for scalar, project overview, prerequisites, first-time setup and commands.
cli-printing-press AGENTS.md
AGENTS.md instructions for mvanhorn/cli-printing-press, covering cli printing press - development conventions, machine vs printed cli, anti-reimplementation, agent-native surface and default: expose; skip rules are exceptions.
kubb AGENTS.md
AGENTS.md instructions for kubb-labs/kubb, covering agents.md, high-level architecture, project structure and commands, repository setup and plugin ecosystem.
oastools copilot-instructions.md
Instructions for erraggy/oastools, covering github copilot instructions for oastools, how to use this file, project overview, specification references and key openapi specification concepts.
oastools AGENTS.md
Instructions for erraggy/oastools, covering agents.md, project overview, dev environment setup, quick commands and build and test workflow (run after every code change).
oastools GEMINI.md
Instructions for erraggy/oastools, covering gemini.md, project overview, building and running, key commands and development conventions.