wiki-instruction-guard

What problem does it solve?

Prevents agents from executing dangerous, obfuscated, or socially engineered instructions copied from wiki pages by scanning extracted executable content and hard-blocking high-risk patterns before execution.

Core Features & Use Cases

• Deterministic static scanning of fenced code blocks, inline commands, and prose to detect destructive filesystem, exfiltration, git, privilege escalation, obfuscation, and other risky patterns.
• Non-overridable guard rules that unconditionally block social-engineering attempts to disable safety checks and require explicit human confirmation for most blocked findings.
• Domain allowlist opt-in and audit trail: supports a local domain-allowlist for narrowly scoped curl-pipe confirmations and appends all scan outcomes to an append-only audit log for post-incident review.
• Use Case: When an agent fetches a hosted wiki setup page containing curl | sh snippets or installation scripts, run this Skill to detect and block destructive commands and request human approval where appropriate.

Quick Start

Before executing any commands gathered from a wiki page, run the wiki-instruction-guard to scan the page and report blocked or warned patterns.