Agent skills that debug production errors, add error tracking to an app, and check a release after a deploy. They use the Agent Skills format, so they work in Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and other agents that read skills. Each skill reads LogBrew through the hosted MCP server or the logbrew CLI. With no tool connected, each skill follows the same steps with the stack tr
npx mdskills install LogBrewCo/logbrewco-skillsRelated
1# LogBrew Skills23Agent skills that debug production errors, add error tracking to an app, and4check a release after a deploy. They use the5[Agent Skills](https://agentskills.io) format, so they work in Claude Code,6Codex, Cursor, GitHub Copilot, Gemini CLI, and other agents that read skills.78Each skill reads [LogBrew](https://logbrew.co) through the hosted MCP server9or the `logbrew` CLI. With no tool connected, each skill follows the same10steps with the stack traces, logs, and counts that you give it.1112| Skill | What it does | Ask your agent |13| --- | --- | --- |14| [`debug-production`](plugins/logbrew/skills/debug-production/SKILL.md) | Finds the failing issue, reads its stack and first-seen release, pulls the logs and trace around the first event, checks who is affected, writes the fix, and checks that new events stop. | "Why did checkout break?" |15| [`instrument-app`](plugins/logbrew/skills/instrument-app/SKILL.md) | Detects the framework, installs the matching LogBrew SDK, sets service, environment, and release, sends a test event, and reads it back. Includes logging rules that help with any tool. | "Add error tracking to this app." |16| [`verify-release`](plugins/logbrew/skills/verify-release/SKILL.md) | Compares the issues, error logs, and trace error rate of a new release with the previous release, and decides go, rollback, or wait. | "Is the deploy safe?" |1718## Install1920Any agent that reads skills:2122```bash23npx skills add LogBrewCo/skills24```2526Claude Code, with the skills and the LogBrew MCP server in 1 plugin:2728```text29/plugin install logbrew --marketplace LogBrewCo/skills30```3132On Claude Code versions before 2.1.275, add the marketplace first:3334```text35/plugin marketplace add LogBrewCo/skills36/plugin install logbrew@logbrew37```3839Or copy a folder from [`plugins/logbrew/skills/`](plugins/logbrew/skills) into40the skills folder of your agent. Each folder works alone.4142## Connect LogBrew4344Use 1 of these:4546- **MCP server.** The Claude Code plugin adds it. In other clients, add47 `https://mcp.logbrew.co/mcp` as a remote (HTTP) MCP server and sign in.48 The connection reads only the projects that you approve. See49 [Connect an AI agent with MCP](https://docs.logbrew.co/guides/mcp).50- **CLI.** Install it with `brew install LogBrewCo/tap/logbrew` or51 `npm install -g logbrew-cli@latest`, then run `logbrew login`. See52 [Install and use the LogBrew CLI](https://docs.logbrew.co/guides/cli).5354Free plan: 50 GB of logs a month, no card.55[Create an account](https://logbrew.co).5657## What is in each skill5859```text60plugins/logbrew/skills/61 debug-production/62 SKILL.md63 references/cli.md CLI commands for each step64 references/mcp.md MCP operations for each step65 references/worked-example.md A full investigation66 instrument-app/67 SKILL.md68 references/frameworks.md Install commands and setup code per framework69 references/logging-rules.md Logging rules for any tool70 verify-release/71 SKILL.md72 references/operations.md CLI commands and MCP operations73```7475## Checks7677`python3 scripts/check.py` checks the skill format, the links, and the eval78cases in [`tests/evals.json`](tests/evals.json). It also checks that every79`logbrew` command matches the help of the installed CLI, and that every MCP80operation and input field is in the published operation catalog. It makes no81model calls. CI runs it against the newest CLI release on every pull request.8283Each eval case is a prompt with the steps, commands, and operations that the84skill must give for it. To test a skill with a model, give the prompt to your85agent with the skill installed and compare its steps with the case.8687## License8889[MIT](LICENSE)90
Full transparency — inspect the skill content before installing.