webflow-incident-runbook

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches service status information from the official Webflow status API and user-defined application health endpoints.- [COMMAND_EXECUTION]: Instructions are provided for managing infrastructure secrets and environment variables using external CLI tools such as kubectl, vercel, and fly.- [COMMAND_EXECUTION]: The skill executes local file system operations, including reading application logs from /var/log/app/ and running a local diagnostic script named webflow-debug-bundle.sh.- [PROMPT_INJECTION]: An indirect prompt injection surface is present.
  • Ingestion points: External data is ingested from status.webflow.com and an internal health endpoint in SKILL.md.
  • Boundary markers: No delimiters or ignore-instructions are used when interpolating external data.
  • Capability inventory: The agent has access to Bash and python3 for executing logic derived from these inputs.
  • Sanitization: Data from external JSON responses is printed directly to the output without filtering or escaping.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 04:42 PM