gameflow-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Oct 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell commands and local scripts to interact with the GameFlow API.
  • Evidence: scripts/diagnose.sh is provided to collect identity, game configuration, and server status via curl.
  • Evidence: Multiple curl examples are used to check API connectivity and fetch diagnostic data from api.gameflow.gg.
  • Observation: The implementation follows security best practices by reading credentials from environment variables ($GAMEFLOW_API_KEY) and using curl -K - to pass sensitive headers via stdin, preventing credential exposure in process lists.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of API error messages and server logs to provide troubleshooting advice.
  • Ingestion points: The agent is instructed to read the exact error body from API responses and fetch server/build logs (SKILL.md, references/diagnostics-api.md).
  • Boundary markers: The skill relies on matching symptoms against a pre-defined references/error-catalog.md to mitigate the risk of arbitrary instruction following.
  • Capability inventory: The agent has the capability to execute curl commands and suggests administrative actions like promoting builds or stopping servers (with user confirmation).
  • Sanitization: Instructions guide the agent to match specific error strings rather than executing content directly from the logs.
  • [SAFE]: All external communication and references target the vendor's official domain (gameflow.gg) and API endpoints (api.gameflow.gg).
  • Evidence: The identity check, diagnostics, and metrics collection are performed against official platform infrastructure.
  • Evidence: The author 'gameflowgg' is consistent with the targeted platform services.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 29, 2026, 04:02 AM