The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read, write, and search external data (contacts, companies, leads, opportunities, activities, and pipelines) from Microsoft Outlook. This creates a surface for indirect prompt injection if the ingested data contains instructions that could influence the agent's behavior.
Ingestion points: Data retrieved from Microsoft Outlook via apideck.crm.contacts.list and other CRM unified API endpoints.
Boundary markers: No specific delimiters or instruction-isolation warnings are mentioned in the documentation for handling external content.
Capability inventory: The skill allows for network operations to Apideck's infrastructure (unify.apideck.com) and supports broad CRM operations (read/write).
Sanitization: No explicit sanitization or filtering logic for external data is documented within the skill files.
[SAFE]: The skill correctly instructs the use of environment variables (process.env.APIDECK_API_KEY, process.env.APIDECK_APP_ID) for secret management rather than hardcoding credentials.
[SAFE]: All external references, including documentation links (learn.microsoft.com) and API endpoints (unify.apideck.com, specs.apideck.com), point to well-known services or official vendor infrastructure.

microsoft-outlook