Instruction integrity
Detects common attempts to override higher-priority prompts or reveal hidden instructions.
Loading…
Private static analysis
Scan an AI agent SKILL.md for prompt injection, destructive commands, secret access, remote shell execution, obfuscation and excessive privileges.
Analysis runs in this browser. No public result URL is generated, and exported reports omit the analyzed source.
Coverage
Detects common attempts to override higher-priority prompts or reveal hidden instructions.
Surfaces destructive filesystem commands, privilege escalation and security-control changes.
Looks for instructions that read, display, collect or transmit credentials and environment files.
Flags remote code piped into shells, obfuscation and referenced external domains.
FAQ
No. It performs deterministic static analysis in the browser and never runs commands or follows instructions from the pasted file.
No. Analysis occurs in the current browser. Content is saved only if you explicitly choose local browser storage or download a report.
No. Static heuristics can miss dynamic behavior. Review referenced scripts, permissions, network destinations and runtime behavior in an isolated environment.
Unexplained credential access, destructive defaults, remote code piped to a shell, hidden payloads and attempts to override system or developer instructions require resolution before installation.