gus-cli

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute sf CLI commands (e.g., sf data query, sf data create record, sf data update record) to interact with a Salesforce org. While these are active commands, the skill explicitly mandates user confirmation for any 'write' actions (create/update) and limits 'read' actions to querying Salesforce metadata and records.
  • [DATA_EXPOSURE]: The skill manages a local cache file at $HOME/.claude/runner-identity.json to store the user's Salesforce identity (userId, username, githubLogin, etc.). This is a standard practice for CLI-based tooling to avoid redundant lookups and does not involve exfiltration of sensitive system credentials like SSH keys or AWS configs.
  • [SAFE]: All external URLs and resources belong to official Salesforce infrastructure (e.g., gus.lightning.force.com) or the author's own organization (forcedotcom).
  • [SAFE]: The skill uses !alias list patterns (specifically sf alias list) within its instructional logic to determine the execution environment, which is a legitimate use case for a CLI-integrated skill.
  • [SAFE]: Instructions for creating work items require a minimum of 20 characters for the description, which is a functional requirement of the target Salesforce system and not a bypass or injection technique.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 10:06 AM
Security Audit — agent-trust-hub — gus-cli