Skill Seekers Skill Builder
Turn documentation, repositories, and PDFs into AI-ready skills while detecting conflicts between documentation and implementation.
The skill names scraping, enhancement, upload, and installation tools that may create external effects, but it requires no user confirmation and does not disclose detailed data flows, sensitive-data handling, permission boundaries, rollback, or pre-upload review. The README mentions API keys and external services but does not complete these controls, so points are deducted. No malware, credential theft, covert exfiltration, or destructive default was evidenced.
SKILL.md gives a source-type routing table and staged workflow, while the repository includes test configuration and Docker test workflows, providing some reproducibility signals. However, tool availability, required parameters, abnormal-input behavior, diagnostic errors, and recovery paths are unspecified. The claim of 40 MCP tools also conflicts with the Docker workflow's 25-tool description, limiting the score under static review.
Audience, use cases, input patterns, and major source types are clear, covering documentation, GitHub, PDFs, videos, and local directories; the repository also declares Simplified Chinese documentation and international support. Non-fit boundaries, trigger precision, output contracts, and mainland-China network reachability are not established, and the core flow may depend on external sites, MCP, and LLM services, so points are deducted.
The skill has concise metadata, trigger guidance, source detection, and a staged workflow. Repository context supplies an MIT license, version metadata, README material, and testing/CI signals. The skill itself lacks installation details, parameter specifications, output examples, FAQs, known limitations, a changelog, and explicit maintenance ownership/update procedure; the inconsistent tool-count description also reduces maintainability.
The workflow covers detection, scraping, enhancement, packaging, and vector-database export, and README commands support the claimed core use case. Static evidence does not verify generated-skill completeness, directly usable output formats, conflict-detection accuracy, or actual benefit over manual alternatives. The score therefore uses the static-review ceiling and deducts for missing outcome verification.
Committed test directories, pytest configuration, and some real Docker CI checks provide auditable evidence. The supplied material does not show that these tests cover the skill-builder MCP paths, nor does it provide third-party execution evidence or independent corroboration. Under the static calibration, the score cannot exceed 5.
- Before scraping, enhancing, uploading, or installing, require explicit confirmation of scope, destinations, network requests, transmitted data, and credentials.
- Do not treat the README's '3700+ tests' or performance/quality claims as validation of this skill path; add tests covering source routing, conflict detection, and failure paths.
- The MCP tool count is inconsistent between 40 and 25; verify the actual server capabilities, dependency versions, and reachability of required external services from mainland-China networks.
What it does & when to use it
Skill Builder is the skill-builder component of the Skill Seekers repository. It creates AI skills from documentation sites, GitHub repositories, PDFs, videos, codebases, and other knowledge sources. It operates through the Skill Seekers MCP server, which provides scraping, analysis, enhancement, packaging, and export tools. It is intended for developers and AI engineers who need structured, reusable context from technical source material.
Detects the source type from a user request and selects tools such as scrape_docs, scrape_github, scrape_pdf, scrape_video, scrape_codebase, or scrape_generic. It can generate and validate configurations, estimate documentation scope, enhance generated skills, package them for target platforms, and export content to Weaviate, Chroma, FAISS, or Qdrant. GitHub workflows can perform AST analysis, API extraction, and conflict detection between documentation and implementation.
- A developer wants to turn a framework's documentation into an AI coding skill.
- A team needs to analyze a GitHub repository and compare documented APIs with the actual implementation.
- A researcher wants to convert a PDF manual or other knowledge source into structured LLM context.
- An AI engineer needs to export processed knowledge to Chroma, FAISS, Qdrant, or Weaviate.
Pros & cons
- Supports documentation websites, GitHub repositories, PDFs, videos, codebases, and other source types.
- Provides AST analysis, API extraction, and documentation-versus-code conflict detection.
- Supports AI enhancement, platform packaging, and multiple vector-database export targets.
- Exposes the workflow through natural-language MCP tool calls.
- Requires installation and configuration of the Skill Seekers MCP server for the documented workflow.
- Some source types require optional dependencies, including video, Jupyter, PPTX, and Notion support.
- Online scraping requires network access and GitHub workflows may encounter API rate limits.
- The supplied SKILL.md does not document the MCP server endpoint, authentication flow, or complete parameters for every tool.
How to install
Install Skill Seekers with pip install skill-seekers. For the MCP server, install pip install skill-seekers[mcp] and run python -m skill_seekers.mcp.server_fastmcp; the repository also provides ./setup_mcp.sh for agent configuration. The repository requires Python 3.10 or newer.
How to use
In a compatible client, make a request such as Create an AI skill from facebook/react, Convert this PDF into an LLM-ready skill, or Update the existing skill and detect documentation/code conflicts. The skill selects the appropriate scraper, after which you can request enhancement, packaging, or vector-database export.