The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill retrieves and exposes sensitive credentials. In the SWARMING and SSO checklists within references/configuration-checklists.md, the agent is instructed to use SOQL to fetch the ConsumerSecret (Azure Client Secret) from the AuthProvider object. SKILL.md and references/report-generation.md require this secret to be displayed in cleartext in the 'Configured Value' column of the final validation report.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection via unsanitized user inputs.
Ingestion points: The skill collects usernames, Azure AD Tenant IDs, and Experience Site URLs in SKILL.md (Step 1).
Boundary markers: No delimiters or instruction-bypass warnings are used for these inputs.
Capability inventory: The skill uses the headless-360 toolset (dispatch and dispatch_readonly) to perform SOQL queries and REST API requests. Inputs are directly interpolated into SOQL strings (e.g., AssigneeId = '<userId>') and URL paths (e.g., .../organization/{tenantId}/extensions).
Sanitization: The skill lacks instructions to sanitize or validate these inputs before execution, allowing potentially malicious content to influence backend queries.

service-itsm-teams-debug