Launch, monitor, pull, and cancel hosted Beaker optimization runs via the beaker CLI.
npx skills add https://github.com/rilixai/beaker-skill --skill beaker-usage-rilixai
Please help me install this Agent Skill. Skill: beaker-usage Source: https://github.com/rilixai/beaker-skill/tree/main/skills/beaker-usage Command: npx skills add https://github.com/rilixai/beaker-skill --skill beaker-usage-rilixai
beaker run trigger, monitor with bounded polling, pull completed results, and cancel runs safely with explicit authorization. - Launch Discovery: Verify auth and GitHub connection, select the correct agent by matching github_repository, list remote branches and hosted datasets, and check required environment variables before launching. - Model Comparison: Optionally launch comparison runs with 1–8 --optimization-model values drawn from beaker model list --available-only, only when the developer explicitly requests it. - Use Case: A developer asks to "launch a Beaker optimization run on the staging branch with dataset prompts@v3." The skill validates the connection, confirms the branch and dataset, states the exact launch parameters, and triggers the run—then reports the full run ID and UI URL. ## Quick Start Ask your agent to launch a Beaker optimization run for this repository and it will discover the connected agent, dataset, and branch before triggering the hosted run.High-intent search queries and answers about installing and using this skill.
Run `beaker run trigger --integration-id <id> --agent <agent> --dataset <name@revision> --json` after verifying auth, GitHub connection, and required environment variables. Launch only after explicit developer authorization, and capture the full run ID and web URL from the JSON output.
Use `beaker run status <run-id> --watch --poll-interval 15 --poll-timeout 45 --json` for bounded polling. Exit code 3 means the run is still active, 0 means completed, and 1 means terminal but unsuccessful—report meaningful state changes without implying setup is unfinished.
No. Hosted GitHub sources accept remote branches only; tags and commit SHAs passed via `--ref` return a 422 error. Ask for a remote branch pointing to the desired commit, preferably one returned by `beaker github branches`.
Use beaker-usage to operate an already-configured integration: launching, monitoring, pulling results, and cancelling runs. If the config, agent, GitHub association, dataset contract, or integration is missing or broken, stop and use beaker-setup instead of repairing it incidentally.
Repeat `--optimization-model provider:model` 1–8 times on `beaker run trigger`, using only models from `beaker model list --available-only --json`. Do this only when the developer explicitly requests a comparison; the default run optimizes the production system.
Exit code 3 means the run is still active or a bounded poll timed out—it is not a failure. Continue monitoring only if the developer requested it; exit 0 means completed, 1 means terminal but unsuccessful, and 2 means a client or lookup error.
sammorrispbValidate Circle API environment variables and test Admin API connectivity.
home-assistantIdentify Home Assistant pull requests ready to merge by verifying CI, reviews, and conflicts.
jnlanahanIntegrate Anthropic LLM capabilities with model selection, prompt caching, and LangSmith tracing.
finbarrtimbersMonitor Beaker experiments until completion with status, progress, and logs.