Provides a lightweight local CLI wrapper around @sylphx/pdf-reader-mcp to extract text, metadata, page counts, images and table detections from PDFs. Designed for environments where the full MCP runtime is not available, it runs ad-hoc via bunx and returns structured extraction results for summarization.
Trigger this skill when a user asks to extract or summarize the contents of PDF files (local paths or URLs), request metadata, or needs page-level extraction or table detection on resource-constrained devices. Useful in 'pi' style agents that lack native MCP integration.
Works with lightweight local agents and CLIs that can invoke Bun/Node tooling; suitable for pi agents and other local assistants that can run shell scripts.
A CLI wrapper that uses bunx to spawn the Sylphx PDF Reader MCP server and extract text, metadata, page info, images and tables from PDF files. The SKILL.md is well-structured with clear triggers, options, and examples. The script is cleanly written with proper arg parsing, JSON-RPC client, timeouts, and cleanup. Script could not be executed because the runner doesn't handle .mjs extensions (requires bun). Main security concern: bunx downloads and executes a remote package without verification.
bun (required runtime for .mjs script)bunx is similar to npx — downloads and executes remote code without checksum verification. This is a moderate security concern but standard practice in the JS ecosystem. The script itself is well-structured with no malicious patterns. No credentials, no phone-home, no destructive commands. The .mjs extension caused the test runner to skip execution.