How does the skill handle different user permissions?

It processes the user's permission tier (READ_ONLY, WRITE_LOCAL, or FULL_ACCESS) to determine if the requested action is appropriate for their access level.

Is any tool access required for this skill?

No, this is a classification-only skill that works by analyzing text input and does not require access to external tools or the file system.

What is the primary purpose of the Security Gate skill?

It serves as a screening layer for the telclaude bridge, evaluating Telegram messages to prevent malicious or destructive commands from being executed by Claude.

Does it track user behavior over time?

The skill accepts a 'flaggedHistory' input, allowing it to apply stricter scrutiny to users who have previously attempted risky or prohibited actions.

Can it detect prompt injection attempts?

Yes, the skill is specifically designed to identify and BLOCK attempts at prompt injection or efforts to bypass established safety protocols.

Security Gate for Telclaude

Name: Security Gate for Telclaude
Author: avivsinai

byavivsinai

•

Security & Testing

Screens and classifies inbound Telegram commands to protect Claude Code environments from destructive or malicious actions.

The Security Gate skill acts as an intelligent firewall for the telclaude bridge, evaluating user messages before they reach the Claude Code execution environment. It utilizes LLM-based screening to categorize requests into ALLOW, WARN, or BLOCK states based on the user's permission tier and historical behavior. By proactively detecting prompt injections, destructive commands, and unauthorized system access attempts, it ensures developers can safely interact with their codebases via mobile messaging while maintaining robust security boundaries.

Key Features

01Real-time classification of inbound messages into ALLOW, WARN, or BLOCK categories

021 GitHub stars

03Automated detection of prompt injection and safety bypass attempts

04Tiered permission evaluation supporting READ_ONLY, WRITE_LOCAL, and FULL_ACCESS levels

05Structured JSON output providing classification rationale and suggested access tiers

06Context-aware screening that accounts for prior user warnings and history

Use Cases

01Preventing accidental execution of destructive shell commands like 'rm -rf' via Telegram

02Protecting sensitive project files and secrets from unauthorized exfiltration attempts

03Enforcing strict read-only access for external contributors or guest users on the bridge

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add avivsinai/telclaude security-gate

For use in Claude.ai and ChatGPT

Download Skill