plan-perf-audit

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is designed for performance auditing and planning. It explicitly forbids automated fixes, code deletion, or behavior changes through a 'Preservation Contract' (references/preservation-contract.md), ensuring that it remains a diagnostic and planning tool rather than an active execution agent.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external monitoring and diagnostic tools which could theoretically contain malicious instructions.
  • Ingestion points: Performance logs from Supabase, transaction data from Sentry, and CLI output from Lighthouse (SKILL.md, references/audit-scope.md).
  • Boundary markers: The instructions lack explicit text delimiters for data ingestion but use a 'Preservation Contract' to constrain the agent's response logic.
  • Capability inventory: The skill has no file-writing, code-execution, or state-changing capabilities, as it is strictly 'plan-only'.
  • Sanitization: There is no mention of sanitizing or escaping the data retrieved from external logs or diagnostic tools.
  • Context: Because the skill is limited to planning and reporting, and the data sources are typically developer-controlled infrastructure, this represents a low-risk surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 07:41 AM
Security Audit — agent-trust-hub — plan-perf-audit