close
  • English

      • AI

      To help AI better understand Rsdoctor's features, configuration, and best practices so it can provide more accurate assistance during day-to-day development and troubleshooting, Rsdoctor provides the following capabilities:

      Agent Skills

      Agent Skills are domain-specific knowledge packs that can be installed into Agents, enabling them to give more accurate and professional suggestions or perform actions in specific scenarios.

      In the rstackjs/agent-skills repository, there are many skills for the Rstack ecosystem. The skills related to Rsdoctor include:

      • rsdoctor-analysis: Use Rsdoctor for build analysis and provide optimization recommendations.

      In Coding Agents that support skills, you can use the skills package to install a specific skill with the following command:

      npm
      yarn
      pnpm
      bun
      deno
      npx skills add rstackjs/agent-skills --skill rsdoctor-analysis

      After installation, simply use natural language prompts to trigger the skill, for example:

      Use Rsdoctor to analyze this project and provide optimization suggestions

      Agent CLI

      @rsdoctor/agent-cli is a command-line tool for Agents. It reads Rsdoctor build analysis data and outputs structured JSON, making it easier for Agents to search, filter, and analyze the results.

      It is typically used together with the rsdoctor-analysis skill: first use the skill to guide the project through Rsdoctor build analysis, then use Agent CLI to query the generated analysis data.

      • Install it in your project (optional, because the skill can automatically install this dependency globally):
      npm
      yarn
      pnpm
      bun
      deno
      npm add -D @rsdoctor/agent-cli
      • The main rsdoctor-agent commands include:
      rsdoctor-agent --help
rsdoctor-agent --version
rsdoctor-agent list
rsdoctor-agent query <tool-name> --data-file <path> [--input <json>]
rsdoctor-agent <group> <subcommand> --data-file <path>

      Description:

      • rsdoctor-agent --help: Show command help, including available subcommands, argument descriptions, and examples.
      • rsdoctor-agent --version: Show the current @rsdoctor/agent-cli package version.
      • rsdoctor-agent list: List the currently available direct grouped subcommands (for example, group subcommand / group.subcommand) for interactive use. This output is not the same as the catalog <tool-name> values used by query.
      • rsdoctor-agent query <tool-name> --data-file <path> [--input <json>]: Invoke a specified catalog tool and return the result.
        • <tool-name>: The catalog tool name to invoke, for example packages_duplicates.
        • --data-file <path>: Path to the Rsdoctor analysis data file (required).
        • [--input <json>]: Optional extra input passed to the tool (JSON string).
      • rsdoctor-agent <group> <subcommand> --data-file <path>: Run grouped commands directly for interactive analysis.

      Execution examples

      • Skill execution example:

        • Prompt: Use rsdoctor to help me do bundle analysis to see from which aspects the product volume can be optimized.

        As shown in the animation above, the AI tool runs the corresponding @rsdoctor/agent-cli commands based on the Rsdoctor skill, retrieves the required data, and outputs optimization suggestions after consolidating the analysis. It also provides a deeper follow-up analysis plan, so you can follow the guidance to further investigate project build issues.



      • Command example (triggered by skill execution):

      # Build optimize with side-effects pagination tuned for analysis
rsdoctor-agent bundle optimize --data-file ./dist/rsdoctor-data.json --side-effects-page-size 10

# Example: return only required fields
rsdoctor-agent query packages_duplicates --data-file ./dist/rsdoctor-data.json --filter "<command-specific-filter>"

      MCP Server

      Rsdoctor provides MCP Server so AI tools can query your local build analysis data. See the MCP Server documentation.

      llms.txt

      llms.txt is a standard that helps LLMs discover and use project documentation. Rsdoctor follows this standard and publishes the following two files:

      • llms.txt: A structured index file containing the titles, links, and brief descriptions of all documentation pages.
      https://rsdoctor.rs/llms.txt
      • llms-full.txt: A full-content file that concatenates the complete content of every documentation page into a single file.
      https://rsdoctor.rs/llms-full.txt

      You can choose the file that best fits your use case:

      • llms.txt is smaller and consumes fewer tokens, making it suitable for AI to fetch specific pages on demand.
      • llms-full.txt contains the complete documentation content, so AI doesn't need to follow individual links - ideal when you need AI to have a comprehensive understanding of Rsdoctor, though it consumes more tokens and is best used with AI tools that support large context windows.

      Markdown docs

      Every Rsdoctor documentation page has a corresponding .md plain-text version that can be provided directly to AI. On any doc page, you can click "Copy Markdown" or "Copy Markdown Link" under the title to get the Markdown content or link.

      https://rsdoctor.rs/guide/start/intro.md

      Providing the Markdown link or content allows AI to focus on a specific chapter, which is useful for targeted troubleshooting or looking up a particular topic.

      AGENTS.md

      You can create an AGENTS.md file in the root of a project that uses Rsdoctor. This file follows the AGENTS.md specification and provides key project information to Agents.

      Here is an example of Rsdoctor-related content you can add to AGENTS.md:

      # AGENTS.md

You are an expert in JavaScript, Rsdoctor, and build analysis.

## Tools

### Rsdoctor

- Run `RSDOCTOR=true npm run build` to build the app with Rsdoctor

## Docs

- Rsdoctor: https://rsdoctor.rs/llms.txt

      You can also customize it for your project, adding more details about the project structure, overall architecture, and other relevant information so Agents can better understand your project.

      Tip

      If you are using Claude Code, you can create a CLAUDE.md file and reference the AGENTS.md file in it.

      CLAUDE.md
      @AGENTS.md