supabase-reliability-patterns
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements patterns such as an
OfflineQueue(using IndexedDB) and adualWritestrategy (using Redis) to buffer and reconcile data. Because these systems ingest and process data from external storage layers, they create an attack surface for indirect prompt injection if those external stores contain malicious instructions intended to be executed by the agent during data processing. - Evidence:
references/offline-degradation-health-dualwrite.md(theOfflineQueue.flushandreconcilePaymentsmethods process stored records). - [COMMAND_EXECUTION]: The skill metadata requests access to
Bash(supabase:*)andBash(curl:*)tools. While the instructions use these for legitimate tasks like health checks, these tools grant the agent extensive shell-level capabilities that could be misused if the agent is compromised by other attack vectors. - Evidence:
SKILL.md(allowed-tools header).
Audit Metadata