DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to cut off dangerous tool calls from a relay before the client runs them with Think Watch Lite

Stop answers that download code, read credentials or add startup items before Claude Code or Codex executes them.

Remote script. Protection only covers the rules listed in the README, so review the rule list yourself before trusting a relay with untrusted tool use.

Who it is forA developer who uses a relay or third-party model and wants a safety net against malicious tool calls inside its answers.

The problemA relay can rewrite answers, so an answer may carry a tool call that downloads and runs code or reads private keys.

The outcomeAnswers carrying the listed dangerous tool calls are cut off before the client can run them, and each rule can be set to its own action.

What you need first

  • A supported computer: macOS 12 or later on Apple silicon, Windows 10 21H2 or later, or a Linux x86_64 or aarch64 machine
  • At least one upstream you already have access to, such as an API key, a relay, or a local model
  • The AI client you want to connect, such as Claude Code or Codex, already installed

Steps

  1. Install the app and connect Claude Code or Codex to the gateway.
  2. Add the relay or third-party upstream you want to protect.
  3. Open the Dangerous commands protection and review its rules while it is in Monitor mode, which records and notifies.
  4. Turn on the rules you want to enforce, one at a time, and set the action for each rule.
  5. Run a harmless test prompt through the relay and check the request history for the recorded result.

Prompt for your agent

Using the repo ThinkWatchProject/ThinkWatch-Lite, enable the Dangerous commands protection for the upstream <upstream name> and report which rules are in Monitor mode and which are enforcing. Expected output: a list of rules with their action. Verify that a test answer containing a flagged tool call is recorded in the request history.

Caveat. The README lists which kinds of tool calls are covered, so other risky commands may pass through unless you add a rule for them.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page