planetscale-change-gates-and-approval-contract
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a policy for processing database mutations (DDL, DML, and configuration changes) that may originate from untrusted data sources or LLM-generated proposals. This defines the primary attack surface for indirect prompt injection via malicious change sets.
- Ingestion points: Operations described in Class B, C, and D, such as schema recommendations, telemetry-generated reports, and deploy requests (SKILL.md).
- Boundary markers: The skill mandates strict approval requirements including explicit target confirmation and human review, which act as high-integrity boundaries against accidental execution.
- Capability inventory: The governed environment includes high-privilege actions like production DDL/DML, credential rotation, and network posture changes.
- Sanitization: Enforces pre-execution validation plans, rollback strategies, and exact command confirmation to mitigate the risk of malicious instructions hidden within proposed changes.
Audit Metadata