skills/spotdraft/oncall-investigator-skills/executed-contract-pending-pre-send-approval/Gen Agent Trust Hub
executed-contract-pending-pre-send-approval
Pass
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is an instructional runbook that provides diagnostic SQL queries and identifies internal code paths relevant to a specific data-state issue. No malicious patterns, unauthorized data exfiltration, or remote code execution vectors were detected. The skill aligns with its stated purpose of assisting in contract status troubleshooting.
- [PROMPT_INJECTION]: Analysis of indirect prompt injection surface (Category 8): The skill ingests untrusted data through placeholders such as
{workspace_id}and{contract_id}within SQL query templates. While these are necessary for the skill's diagnostic function, they represent an entry point for untrusted data into the agent's workflow. - Ingestion points:
{workspace_id},{cluster},{contract_id}provided by the user/agent. - Boundary markers: No explicit delimiters or 'ignore' warnings are present within the SQL code blocks.
- Capability inventory: The skill involves executing read-only BigQuery SQL statements and suggests (but does not execute) Django shell mitigation scripts.
- Sanitization: No sanitization logic is provided in the instructions; the agent is responsible for validating inputs before interpolating them into the SQL templates. This surface is considered safe given the intended administrative/support context.
Audit Metadata