troubleshoot-app

Warn

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is instructed to read environment variables (e.g., .env*), deployment metadata, and repository configuration to identify live data sources. It also performs read-only queries against external systems including authentication (Clerk, Auth0), billing (Stripe, Paddle), and databases (Convex, Supabase, etc.).\n
  • Evidence: Read project docs, .env* variable names, scripts, package files, and deployment metadata to identify systems in use.\n
  • Mitigation: Explicit instructions are provided to redact secrets, tokens, and personal data from all summaries and responses.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from several sources that could contain malicious instructions designed to influence the agent's behavior.\n
  • Ingestion points: GitHub issue bodies, PR descriptions, application logs, analytics events, and data-plane provider responses.\n
  • Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands in the processed data.\n
  • Capability inventory: The skill has the ability to modify the local repository (applying fixes), execute shell commands via CLIs, and spawn subagents.\n
  • Sanitization: While the skill enforces redaction of sensitive data, it does not include logic to filter or sanitize natural language prompt injection attempts in the ingested content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 23, 2026, 01:31 AM