The Agent Skills Directory

[SAFE]: The skill demonstrates secure credential handling by storing the Notion integration secret in a configuration file (~/.config/notion/token) and explicitly setting file permissions to 0600. This ensures that the token is only readable by the current user.
[EXTERNAL_DOWNLOADS]: Network operations are confined to the legitimate Notion API endpoint (api.notion.com). No requests are made to unknown or untrusted third-party domains.
[COMMAND_EXECUTION]: The skill uses a collection of Ruby scripts that rely exclusively on the language's standard library. By avoiding external gems, the skill minimizes the risk of supply chain vulnerabilities and malicious dependencies.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection, common in tools that retrieve external content.
Ingestion points: scripts/page.rb, scripts/query_database.rb, scripts/blocks.rb, and scripts/search.rb retrieve data from user-controlled Notion pages.
Boundary markers: The scripts do not use explicit delimiters (like XML tags or triple backticks) to wrap the content retrieved from the API when presenting it to the agent.
Capability inventory: The skill includes scripts for modifying Notion data (create_page.rb, update_page.rb, delete_block.rb), which could be targeted by instructions hidden in fetched content.
Sanitization: The retrieved content is displayed to the agent in its raw plain-text form without sanitization for embedded command patterns.

notion