An agent-native PDF operations skill: inspect, edit, fill, transform, secure, extract, and validate PDF documents through one coherent CLI and structured API.
PDF handling is expressed through document intent, not low-level PDF manipulation commands. The toolkit inspects a document, classifies the job, routes to the right engine, executes, and validates the output — returning structured results, warnings, and machine-readable manifests for every operation.
app/pdf-toolkit/— TypeScript source, tests, fixtures, and build scripts. See its README for development and the full capability list.app/pdf-toolkit/skill/pdf-toolkit/— the distributable agent skill:SKILL.md, the self-contained bundled CLI (cli/pdf.cjs), command/template references, and a ready-to-try example.pdf-toolkit.skill— the packaged skill (zip) for direct installation.
The bundled CLI needs Node 20+ plus external binaries: qpdf and poppler-utils (brew install qpdf poppler on macOS, apt-get install qpdf poppler-utils on Debian/Ubuntu). Ghostscript is optional and improves compression.
node app/pdf-toolkit/skill/pdf-toolkit/cli/pdf.cjs doctor --json
node app/pdf-toolkit/skill/pdf-toolkit/cli/pdf.cjs inspect contract.pdf --jsonAgents should start from SKILL.md, which teaches the inspect → classify → execute → validate workflow.
cd app/pdf-toolkit
npm install
npm run build # bundles dist/pdf.cjs and dist/api.cjs
npm test # 31-scenario end-to-end smoke suiteContributions are welcome — see CONTRIBUTING.md for setup, project layout, and the rules every change must keep (structured results, secret scrubbing, skill-bundle sync). Security reports go through SECURITY.md, not public issues.
MIT — see LICENSE.