iterate-post-launch
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources, creating a surface for indirect prompt injection.
- Ingestion points: Fetches data from Sentry issue descriptions (
sentry:search_issues), production log messages (supabase:query_logs), and web search results (firecrawl:firecrawl_search). It also crawls the live application UI via Playwright. - Boundary markers: The instructions do not define clear delimiters or instruct the agent to ignore instructions embedded within the logs or search results.
- Capability inventory: The agent has significant capabilities including executing SQL commands (
supabase:execute_sql), applying database migrations (supabase:apply_migration), modifying source code, and driving a headless browser via Playwright. - Sanitization: There is no evidence of sanitization or filtering of the ingested external content before it is processed by the agent.
- [EXTERNAL_DOWNLOADS]: The skill downloads and executes the Playwright CLI at runtime.
- Evidence: The command
npx --yes @playwright/cli@latestis used to ensure the latest version of the tool is available for verification tasks. - [COMMAND_EXECUTION]: The skill executes shell commands and database queries as part of its core loop.
- Evidence: Uses Playwright CLI commands (
open,console,requests,screenshot) to interact with the live application and executes SQL via the Supabase MCP to inspect logs and verify fixes.
Audit Metadata