The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The diagram rendering component (scripts/diagram-to-image.ts) dynamically imports and executes remote JavaScript code.\n
Source URLs: https://esm.sh/@excalidraw/excalidraw@0.18.0, https://viewer.diagrams.net/js/viewer-static.min.js\n
Execution Method: Playwright page.evaluate() and page.setContent() within a browser environment.\n- [EXTERNAL_DOWNLOADS]: The skill installs required runtimes and fetches assets from well-known sources.\n
Setup: scripts/render-diagram.sh executes npm install for package.json dependencies and playwright install for the Chromium browser binary.\n
Runtime: Downloads generated diagram source files from the anygen.io vendor domain.\n- [COMMAND_EXECUTION]: The skill coordinates background workflows using sessions_spawn in SKILL.md.\n
Scripts: Runs bundled scripts anygen.py and render-diagram.sh.\n
System Commands: Performs cleanup of temporary files via rm and interacts with the Feishu/Lark API via curl for image delivery.\n- [DATA_EXFILTRATION]: User-provided prompts and files are uploaded to the vendor's infrastructure.\n
Target: www.anygen.io (Stated purpose as a cloud-powered diagramming service).\n
Data: Transmits user input and reference documents after obtaining consent.\n- [PROMPT_INJECTION]: The skill processes data from a remote API that could influence agent behavior, representing an indirect prompt injection surface.\n
Ingestion points: Suggested task parameters and plan descriptions from scripts/anygen.py prepare.\n
Boundary markers: Absent. The instructions specify preserving and relaying the original content of API suggestions to the user.\n
Capability inventory: Subprocess execution via sessions_spawn, file system access in scripts/anygen.py, and network connectivity.\n
Sanitization: Absent. Data from the remote service is processed directly to drive subsequent task logic.

anygen-diagram