camjac251
UserIntelligent tool permission gate & security hooks for AI assistants (Claude Code, Codex, Antigravity; Gemini deprecated). Features AST shell gating, file guards, and safety scanners.
Categories
Indexed Skills (3)
gate-audit
Mine Claude Code session logs for bash commands getting incorrect permission decisions. Find read-only commands that return "ask" when they should be "allow", unknown commands that need gates, and false categorizations. Triggers include "audit gates", "find missing gates", "edge cases", "false ask", "mine sessions".
review
Batch-promote frequently-asked tool-gates patterns to permanent permission rules in settings.json. Lists pending approvals (commands the user approved one-time through the prompt, accumulated across sessions), shows counts and suggested glob patterns, multi-selects to write at user/project/local scope. Use after several sessions to triage what's piled up, or to share patterns across projects. For single-prompt approvals the user should usually click 'Yes, and don't ask again for X' in the CC prompt instead. It covers the in-session case organically since tool-gates defers benign asks. NOT for manual settings.json edits, hooks, env vars, or MCP config (use update-config). NOT for scanning a transcript to build an allowlist from scratch (use fewer-permission-prompts). When the user wants to triage their pending approval queue, batch-promote frequently-asked patterns to permanent rules, share approvals across projects, or audit what's accumulated. Triggers on 'review pending', 'pending approvals', 'tool-gates re
test-gate
Test how tool-gates evaluates any tool invocation (Bash, Read, Write, Edit, Glob, Grep, Skill, MCP) and show the permission decision (allow / ask / deny / defer). Pipes a synthesized JSON payload to the tool-gates binary and formats the result. Useful for verifying a new gate, debugging why a command was blocked, or distinguishing 'tool-gates explicitly asks' from 'tool-gates lets CC handle it'. NOT for approving a pending permission (use tool-gates:review) or editing rules manually (use update-config). When checking whether a command would be allowed, debugging why something was blocked or denied, verifying a permission rule works, simulating a tool call without running it, or confirming the defer-vs-explicit-ask wire decision. Triggers on 'is this allowed', 'test gate', 'why was this blocked', 'why was this denied', 'check permission', 'simulate tool call', 'would this run', 'does tool-gates catch this'.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.