Skip to content

Security: tencildev/tencil

SECURITY.md

Security policy

Reporting a vulnerability

Please report vulnerabilities privately, not in a public issue or pull request:

Include what you found, the Tencil version (npx tencil --version), your OS and browser, and the steps or a proof of concept to reproduce it. We'll confirm receipt within a few working days, keep you posted while we fix it, and credit you in the advisory unless you'd rather not be named.

Supported versions

Fixes go into the latest release of the tencil package. Please check that the problem still exists there before reporting.

Scope

Tencil drives a coding agent that can edit files and run commands on the developer's machine, so anything that lets someone other than the developer reach it matters. Setup runs one daemon per machine, listening on 127.0.0.1:4747. It serves the overlay script (/tencil.js) and a WebSocket bridge that creates agent threads and sends them prompts.

In scope, for example:

  • A web page that isn't the developer's local dev site getting the daemon to start or steer an agent, read pins or threads, or read credentials: bypassing the Origin check, the Host check (DNS rebinding) or the loopback-only rule for clients without an Origin. These checks live in packages/tencil/src/daemon/security.ts.
  • Agent replies or page content that run as script in the overlay (it renders agent output as text).
  • Credentials (~/.config/tencil/credentials.json, T3 Code pairing tokens) leaking to the page, to logs or to other users on the machine.
  • tencil init or tencil setup writing outside the project or the Tencil config folder, or the login service running something other than Tencil.

Out of scope:

  • What the agent does with a prompt the developer sent it. Tencil passes prompts on; the agent's own permission mode (--runtime-mode, full-access by default) decides what it may do.
  • Exposure the developer opted into: --listen on a network address or --allow-origin for another origin. The CLI documents that these widen access.
  • Vulnerabilities in T3 Code, Claude Code, Codex or the Claude Agent SDK themselves; report those to their maintainers.

There aren't any published security advisories