TOKREPO · ARSENAL
Promoted · Safety

Agent Safety Toolchain Pack (MCP · CLI)

Promoted pack for safer tool-calling: prompt-injection defense, replayable automation, deterministic IAM drafts, and precise code indexing for agents.

6 assets

Agent Safety Toolchain Pack (MCP · CLI) is a hand-picked TokRepo arsenal pack — 6 AI assets bundled together so you can install the whole stack with one command. Each asset has been verified live (status=public, current download), tagged with its primary type, and curated to work alongside the others in this pack.

To install, run `tokrepo install pack/agent-safety-debuggable-toolchain-2026-05` in your terminal — or hand the same command to your AI agent (Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Cline, Roo Code, Windsurf are all supported). The TokRepo Node CLI fetches each asset from the TokRepo registry, places skill files into `.claude/skills/`, prompts into your prompt directory, MCP servers into your client config, and surfaces a summary of what was installed.

Without a pack, you would need to discover, vet, and install each of these 6 assets one by one — typically 30 to 60 minutes per asset. The pack collapses that into a single command, and stays current as the underlying assets get re-published. If any asset goes private or is removed, the pack auto-detects it and skips that entry on next install — no broken bundle.

INSTALL · ONE COMMAND
$ tokrepo install pack/agent-safety-debuggable-toolchain-2026-05
hand it to your agent — or paste it in your terminal
What's inside

6 assets in this pack

Prompt#01
Defender — Prompt Injection Defense for Tool Calls

Defender blocks indirect prompt injection hidden in tool outputs; verified 97★ and README claims ~10ms CPU latency with a bundled ~22MB ONNX model.

by Prompt Lab·13 views
$ tokrepo install defender-prompt-injection-defense-for-tool-calls
MCP#02
Beelzebub — LLM Deception Runtime (MCP Honeypot)

Beelzebub deploys LLM-powered decoys across SSH/HTTP/TCP/TELNET/MCP; verified 1,989★ and ships Docker Compose, Helm, and a structured CLI.

by MCP Hub·426 views
$ tokrepo install beelzebub-llm-deception-runtime-mcp-honeypot
MCP#03
flyto-core — Replayable Automation Engine (MCP)

flyto-core traces every automation step and supports replay; verified 270★ and shows a 12-step browser recipe finishing in ~10.3s.

by MCP Hub·11 views
$ tokrepo install flyto-core-replayable-automation-engine-mcp
MCP#04
IAM Policy Autopilot — Baseline IAM via MCP

IAM Policy Autopilot is an MCP server + CLI that analyzes local code to draft baseline IAM policies; verified 357★ and supports Python/Go/TS/JS/Java.

by MCP Hub·7 views
$ tokrepo install iam-policy-autopilot-baseline-iam-via-mcp
Script#05
ast-index — Fast Code Index CLI (Claude Plugin)

ast-index builds an AST code index for fast search/usages/callers; verified 347★ and ships Homebrew + Winget installers and a Claude Code plugin flow.

by Script Depot·1 views
$ tokrepo install ast-index-fast-code-index-cli-claude-plugin
Script#06
altimate-code — Data Engineering AI Harness CLI

altimate-code is a data engineering AI harness with 100+ deterministic tools; verified 582★ and installs via `npm -g` with a `/connect` flow.

by Script Depot·8 views
$ tokrepo install altimate-code-data-engineering-ai-harness-cli
MORE FROM THE ARSENAL

12 packs · 80+ hand-picked assets

Browse every curated bundle on the home page

Back to all packs