The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill integrates with external LabArchives REST-like APIs and processes electronic lab notebook (ELN) entries, XML data, and Inventory v1 JSON schemas, as well as local LA container ZIP files.\n
Ingestion points: The script scripts/notebook_operations.py parses ZIP archives and XML manifests (lamanifest.xml). The agent instructions also direct it to process and interpret data returned from regional ELN and Inventory API endpoints.\n
Boundary markers: SKILL.md contains explicit instructions and warnings to treat all notebook content (text, captions, comments, and URLs) as untrusted data and specifically instructs the agent never to execute instructions found within that content.\n
Capability inventory: The bundled Python scripts (entry_operations.py, notebook_operations.py, setup_config.py) perform local calculations, validation, and file writes, but do not perform network operations directly. However, the agent's environment typically grants network access for the required API calls.\n
Sanitization: scripts/notebook_operations.py implements multiple security checks, including member count limits, compression ratio validation, and strict path-traversal checks that reject absolute paths, .. segments, and backslashes in archive member names.

labarchive-integration