The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads sensitive credentials from the org, specifically the ConsumerSecret from the AuthProvider object, to validate the integration with Microsoft services.
[COMMAND_EXECUTION]: The skill includes instructions to perform a DELETE operation on NamedCredential records via the dispatch tool. This is a state-changing action intended to clear stale authentication tokens and requires explicit user confirmation.
[EXTERNAL_DOWNLOADS]: The skill initiates network requests to well-known Microsoft services (login.microsoftonline.com and graph.microsoft.com) to verify that the Azure AD application configuration matches the Salesforce setup.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user inputs, including usernames, Azure Tenant IDs, and Experience Site URLs, which are then interpolated into SOQL queries and API requests.
Ingestion points: User-provided inputs (username, Tenant ID, Site URL) collected in Step 1 of SKILL.md.
Boundary markers: None identified in the prompt interpolation logic.
Capability inventory: The skill uses dispatch_readonly for data retrieval and dispatch for credential deletion in references/report-generation.md.
Sanitization: There is no explicit mention of sanitization or escaping for user-provided identifiers before they are used in queries.

service-itsm-teams-debug