Works with the AI Agents you already use
Pagar: Agentic Tool-Call Firewall & Rollback Engine
Pre-execution tool-call firewall, cloud credential scanner, and atomic blast-radius workspace rollback engine in pure Rust (#![forbid(unsafe_code)]). Blocks destructive shell/SQL/Git commands, detects leaked API keys and prompt injections, and byte-for-byte restores unauthorized file drift.
$29
Pagar: Agentic Tool-Call Firewall & Rollback Engine
Example session with this skill installed
Please lock our workspace scope to ['src/auth.rs'], inspect the command 'git push --force origin main' before running it, and audit the workspace for any out-of-scope drift with auto-rollback enabled.
- Read your context and instructions
- Compiled the pagar
Security Assessment & Firewall Verdict
The pre-execution firewall evaluated the pending tool-call, intercepted a destructive command, and verified the locked workspace blast-radius scope. Execution was halted with a cryptographically signed verdict receipt.
| Security Inspection Gate | Result | Matched Pattern / Policy | Action Taken |
|---|---|---|---|
| Workspace Scope Lock | LOCKED | Authorized: ['src/auth.rs'] (1 file) | Baseline Snapshot Saved |
| Pre-Execution Firewall | BLOCK | PRO-CMD:git push --force | Execution Intercepted |
| Shannon Entropy Check | 3,412 mbits | Normal command entropy | Evaluated |
| Out-of-Scope File Drift | REVERTED | Rogue creation: temp_auth.sh | Atomically Rolled Back |
| Active License Tier | ENTERPRISE_PRO_V1 | Unsealed Cryptographic Vault | Enforced |
{
"action": "Block",
"tier": "ENTERPRISE_PRO_V1",
"matched_rules": [
"PRO-CMD:git push --force"
],
"shannon_entropy_bits": 3412,
"receipt_sha256": "8f3b2a1c9e8d7f6a5b4c3d2e1f0a9b8c7d6e5f4a3b2c1d0e9f8a7b6c5d4e3f21",
"scope_audit": {
"clean": true,
"authorized_modified": ["src/auth.rs"],
"unauthorized_created": ["temp_auth.sh"],
"rolled_back_files": ["temp_auth.sh"],
"state_hash_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
}
}
Next Steps
- The destructive force-push command was blocked; no remote history was overwritten.
- The unauthorized file
temp_auth.shwas deleted automatically;src/auth.rsremains preserved.
Connects securely to your tools. The creator never sees your data.
About this skill
Overview
Pagar is a zero-copy, #![forbid(unsafe_code)] Rust security engine and native Model Context Protocol (MCP) server. It operates directly at the host workspace boundary, intercepting agent tool-calls before execution, scanning outbound payloads for leaked secrets and prompt injections, and enforcing byte-exact atomic rollbacks if an agent mutates unauthorized files.
Core Capabilities
- Pre-Execution Command Firewall: Inspects shell, SQL, and Git commands prior to execution, blocking destructive mutations (
git push --force,rm -rf,DROP TABLE, reverse shells) with deterministic SHA-256 verdict receipts. - Secret Exfiltration & Injection Scanner: Uses Shannon entropy analysis and heuristic pattern matching to detect leaked cloud credentials (AWS, OpenAI, Anthropic, Stripe, GitHub, database URIs) and indirect prompt injection attempts.
- Atomic Blast-Radius Scope Lock & Rollback: Snapshots declared workspace files (
pagar_lock_scope) and atomically restores unauthorized file modifications, deletions, or rogue file creations byte-for-byte (pagar_audit_and_rollback).
Enterprise Pro Commercial License Included
This bundle includes the pre-configured PAGAR_LICENSE_KEY inside mcp.json that cryptographically unseals the embedded ENTERPRISE_PRO_V1 threat vault (SHA256-CTR + MAC), enabling:
- 1,000,000 tool inspections per session (vs. 15 in Community Evaluation)
- 10,000 locked workspace files in atomic blast-radius scope locks (vs. 2 in Community Evaluation)
- 22 Destructive Command Signatures across Shell, SQL, Git, and PowerShell
- 12 Cloud & Database Credential Rules
- 8 Indirect Prompt Injection Signatures
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
Binary is precompiled for Windows x86_64 only. Requires binary to be added to system PATH manually. Limited to 1,000,000 tool-call inspections per session.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 12 days ago
Permissions
File Scopes