team-pipeline

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from Salesforce records (Users, Opportunities) and potentially Slack messages, creating a surface for indirect instructions to be ingested into the agent context.
  • Ingestion points: Data is ingested through multiple dispatch_readonly calls to Salesforce GraphQL/SOQL endpoints (specified in SKILL.md steps 1, 2, and 3) and through Slack search tool calls (step 3b).
  • Boundary markers: The instructions do not define explicit delimiters (like XML tags or unique markers) to isolate untrusted data from the agent's instructions, though it does mandate record linking and exact citations.
  • Capability inventory: The skill possesses the capability to query Salesforce data, perform Slack searches, and render custom UI widgets using the display_widget tool.
  • Sanitization: The skill requires grounding API names to human-readable labels, which provides a layer of structural validation for the data displayed to the user.
  • [COMMAND_EXECUTION]: The skill uses structured API calls to interact with the Salesforce platform.
  • Evidence: The skill body contains specific GraphQL queries and SOQL commands used with the dispatch_readonly tool to retrieve pipeline data, such as closed revenue and forecast aggregates.
  • [SAFE]: The skill is authored by the platform vendor and operates within the expected scope of Salesforce business reporting. Analysis confirms that all network operations target vendor-owned infrastructure and no malicious obfuscation or persistence mechanisms are present.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 06:39 PM