ln-25-persistence-auditor

Pass

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data such as query logs, application metrics, and source code. This creates a surface for indirect prompt injection if those data sources contain adversarial content intended to influence the agent's audit results.
  • Ingestion points: repository files, query logs, metrics, traces, and profiles (specified in SKILL.md).
  • Boundary markers: None explicitly defined to separate untrusted log/trace data from instructions.
  • Capability inventory: Native file search, language server tools, and diagnostic commands (specified in Tool Routing section).
  • Sanitization: No explicit sanitization of ingested logs or code is mentioned in the instructions.
  • [COMMAND_EXECUTION]: The skill references the use of diagnostic tools like EXPLAIN ANALYZE, profilers, and repository-defined tests. However, it includes explicit safeguards requiring these to be run only on non-production, disposable targets and limits actions to read-only queries.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 20, 2026, 04:08 AM
Security Audit — agent-trust-hub — ln-25-persistence-auditor