debug-fe-be-integration

Pass

Audited by Gen Agent Trust Hub on Jul 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: Utilizes the Supabase MCP tool (execute_sql) to verify database state, allowing the agent to run SQL queries based on identified integration issues.
  • [EXTERNAL_DOWNLOADS]: Fetches web content and documentation using the Firecrawl (firecrawl_search) and Context7 (resolve-library-id) MCP tools to research debugging patterns.
  • [PROMPT_INJECTION]: Contains an indirect prompt injection surface as it processes untrusted external data.
  • Ingestion points: Reads production error messages from Sentry and local backend logs from terminal output files.
  • Boundary markers: Absent. The skill does not implement delimiters or explicit warnings to ignore instructions embedded within the processed logs or search results.
  • Capability inventory: The agent can perform database operations via Supabase SQL execution and file system analysis using Grep.
  • Sanitization: Absent. Log data and error context are ingested directly into the analysis workflow without filtering or sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 23, 2026, 07:25 AM
Security Audit — agent-trust-hub — debug-fe-be-integration