managing-path-cleaning-rules
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection because it instructs the agent to process
$pathnamedata which originates from external website visitors. A malicious user could visit a URL designed to look like an instruction.\n - Ingestion points: SQL query results from the
eventstable inSKILL.md(Workflow Step 2).\n - Boundary markers: Absent.\n
- Capability inventory:
execute-sqlandproject-settings-updatetools.\n - Sanitization: Absent.\n- [SAFE]: The skill uses platform management tools (
execute-sql,project-settings-update) for their intended purposes within the environment.\n- [SAFE]: External references to documentation for regex syntax point to well-known service repositories.
Audit Metadata