Skip to content

Research Tools

The executor-research skill packages the discovery stack used for this repo. It lives at .agents/skills/executor-research.

NeedPreferred tool
Search the open web with clean snippetsExa or Firecrawl search through Executor.
Scrape a known page or documentation URLFirecrawl scrape, or Markit for simple URL-to-Markdown conversion.
Discover pages inside a large docs siteFirecrawl map, then scrape selected URLs. Use Webpull when the useful artifact is a local Markdown corpus.
Search GitHub repositories and codeGitHub tools through Executor or the GitHub connector.
Search indexed repositories and docsNia before one-off web fetching, when the source is indexed or can be indexed.
Pull a public docs site into local Markdown filesWebpull with a bounded --max page count and output under .webpull/.
Convert PDFs, DOCX, PPTX, XLSX, HTML, images, audio, ZIPs, or GitHub URLs to MarkdownMarkit.
Classify and extract local PDFs to Markdownpdf-inspector, especially before deciding whether OCR is needed.
Compare broad current claimsExa, Perplexity, or Parallel Search, with source checking before the result is treated as evidence.
Terminal window
npx add-mcp http://127.0.0.1:4789/mcp --transport http --name executor --header 'Authorization: Basic <redacted>'
  • Nia API credentials are present locally, but the bundled scripts may need to be invoked through bash scripts/<name>.sh if execute bits are missing.
  • Markit is installed as markit 0.5.0.
  • Webpull is available through bunx webpull. Its core command shape is bunx webpull <url> -o .webpull/<hostname> -m <pages>.
  • Local PDFs can be classified with npx --yes @firecrawl/pdf-inspector detect <file> --json.
  • Local PDFs can be extracted with npx --yes @firecrawl/pdf-inspector <file> -o .firecrawl/pdf-inspector/<name>.md. Keep full extracted text in ignored research folders unless the source license allows publishing it.
  • Executor source availability should be checked in each session because connector configuration can change.