mlclaw
huggingface/mlclaw/AGENTS.md
These instructions apply to this repository. This repository follows the TypeScript standards enforced by Slophammer. Before finishing a change, run formatting, lint, typecheck, tests, coverage, build, secret scanning, package validation, and both Slophammer checks. Do not run mutation testing during ordinary implementation work. Keep its configuration and CI declaration current so a dedicated mutation run can be performed separately. Do not report work as complete until the original user-visible workflow has been reproduced and verified end to end in its…
AGENTS.md10 starsChanged 2 months ago
- Installs packages
# AGENTS.md These instructions apply to this repository. ## Quality Gates This repository follows the TypeScript standards enforced by Slophammer. Before finishing a change, run formatting, lint, typecheck, tests, coverage, build, secret scanning, package validation, and both Slophammer checks. Do not run mutation testing during ordinary implementation work. Keep its configuration and CI declaration current so a dedicated mutation run can be performed separately. Do not report work as complete until the original user-visible workflow has been reproduced and verified end to end in its real target environment. Tests, mocks, CI, health checks, successful builds, and deployments are supporting evidence, not substitutes. If full verification is impossible, state that clearly and report the work as unverified. ## Release Versions - Treat `package.json` as the only hand-edited source for the MLClaw package, OpenClaw, unYOLO plugin, HF Broker binary, and runtime image versions. - Keep the `openclaw-unyolo` dependency exactly equal to `config.unyoloPluginVersion`. - Do not hand-edit Dockerfile release defaults or `src/mlclaw/release-config.generated.ts`. After editing release metadata, run `npm install --package-lock-only` if dependency metadata changed, then run `npm run release:sync`. - Run `npm run release:check` before committing. CI rejects version drift. ## Runtime Boundary - Treat the OpenClaw process and agent account as untrusted except for credentials that OpenClaw must own to provide a native authenticated transport. - The managed native OpenAI OAuth profile is an explicit trust-boundary exception: OpenClaw may persist and refresh that credential only in its private auth store. It must not receive the deployment encryption key or unrelated backend credentials. - Keep Hugging Face, broker operator, session, encryption, and all other OAuth secrets in the trusted backend only. - Telegram deployments must use separate conversation and unYOLO approval bots in one private user chat. Keep the approval bot token out of OpenClaw's environment and run only one `getUpdates` poller for it. - Browser routes may call only fixed typed broker operations. - Do not add arbitrary HTTP forwarding to broker, Hub, or operator APIs. - Validate unknown file, network, OAuth, and broker payloads at their boundary. - Generated runtime bundles and control UI assets must match their source. - Consume only unYOLO's version 1 wire APIs, persisted state formats, plan schemas, manifests, and protocols. Do not synthesize broker-owned state files or add compatibility behavior; let each broker initialize its own v1 state after coordinated contract changes. ## TypeScript - Keep strict compiler settings enabled. - Do not add explicit `any` or unsafe type operations. - Keep functions below the configured complexity limit. - Add focused tests for every security or lifecycle behavior change.
Discussion
Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.
Posts are public.Sign in to post
No one has posted yet. Be the first.

