The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Document content is retrieved and processed through the entry_describe_ai_parse_content tool (SKILL.md).
Boundary markers: The skill explicitly defines a 'Write Operation Security Rules' section, instructing the agent to never select targets autonomously and always require user confirmation for write actions.
Capability inventory: Includes tools for entry creation (entry_create_entry), content import (entry_import_content), and block updates (block_update_blocks).
Sanitization: Potential risks are managed via behavioral instructions to the agent rather than programmatic sanitization within the scripts.
[DATA_EXPOSURE]:
The sync-folder.ts and upload-files.py scripts access the local file system to compute content hashes and read data for synchronization with the knowledge base.
Configuration instructions in setup.md recommend standard storage of access tokens in environment variables or configuration files (~/.mcporter/mcporter.json).
[REMOTE_CODE_EXECUTION]:
The skill uses standard external libraries such as aiohttp, requests, and typescript to facilitate network requests and script execution. These are standard dependencies for the stated functionality.

lexiang-knowledge-base