The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones a repository from an internal Salesforce source control domain (git.soma.salesforce.com). This infrastructure belongs to the vendor and is considered a well-known service for this context.
[COMMAND_EXECUTION]: The skill uses the sf CLI to deploy Apex and LWC code to a Salesforce organization and executes a local Python script (analytics.py) to serve the dashboard. These operations are consistent with the skill's stated purpose of managing a development dashboard.
[INDIRECT_PROMPT_INJECTION]: The skill ingests session telemetry and interaction data from Data Cloud and Core SOQL tables (e.g., AiAgentInteractionStep__dll, RecActorActionFeed).
Ingestion points: Data Cloud queries via ConnectApi and SOQL queries in SKILL.md.
Boundary markers: None present for the processed data.
Capability inventory: Uses Bash for sf CLI operations and Python execution; Write tool for local configuration.
Sanitization: No explicit sanitization of queried data is mentioned.
Assessment: While the skill processes untrusted interaction data, the context is analytical visualization rather than automated action execution, minimizing the risk of accidental obedience.

sra-analytics