The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Intercom conversations and contacts, which can contain user-generated content, forming a potential surface for indirect prompt injection.
Ingestion points: references/implementation.md (Intercom API response data via client.contacts.find and client.conversations.search).
Boundary markers: No specific boundary markers or 'ignore' instructions are used when processing raw conversation text.
Capability inventory: Performs network requests via fetch, file system writes using fs.writeFile, and database interactions with a localDb object.
Sanitization: Includes a redactIntercomData function designed to mask PII fields such as email, name, and SSN before logging data.
[EXTERNAL_DOWNLOADS]: The skill makes requests to external Intercom API endpoints to retrieve user and conversation data.
Evidence: Network calls to https://api.intercom.io/export/messages/data and related endpoints in references/implementation.md.
[COMMAND_EXECUTION]: The skill provides examples that involve writing data to the local file system using Node.js file system modules.
Evidence: Usage of fs.writeFile in SKILL.md and references/examples.md to persist data bundles.

intercom-data-handling