T3 Code Web App Testing Skill
Launch, retain, and test the T3 Code web app in isolated dev environments, including browser pairing authentication and SQLite state management.
The skill document explicitly instructs to treat pairing URLs as secrets, not to include them in logs or responses, and warns against passing --browser during automated testing to avoid consuming tokens. It teaches isolation via base directories and to avoid touching ~/.t3. The SQLite helper refuses shared home writes by default and backs up before each mutation. There is some implicit user confirmation (e.g., no explicit approval for destructive actions). Therefore, 6 points deducted out of 25.
Instructions are fairly consistent, but key commands (e.g., 'vp run dev', 'node apps/server/src/bin.ts auth pairing create') are unverified in the repo. The doc acknowledges failure cases and offers explicit recovery steps, but static review cannot confirm these commands actually exist and work. Therefore, 12 points deducted out of 20 for unverified dependencies and lack of tested evidence.
The skill is well-defined for a clear audience (agent testing T3 in isolation) and scenarios (start, iterate, recover tokens, inspect DB). It declares boundaries (web only, not mobile) and provides guidance for error conditions. However, it does not mention reachability from mainland China or reliance on overseas services, which is a potential issue for FollowSkills users. Thus, 3 points deducted out of 15.
The document is well-structured with progressive disclosure, reference files, examples, and troubleshooting. It has no versioning or changelog, and maintenance responsibility is not clear. The repo provides MIT license but is early stage ('very very early' with bugs expected). Therefore, 5 points deducted out of 15.
The skill is designed to complete the task of launching, verifying, and testing the T3 app with direct steps. Static review cannot verify its real effectiveness in all cases, and no direct evidence of completing key tasks is provided. Unverified commands also reduce confidence in actual completion. Thus, several points deducted, limited by static constraints.
The repo contains CI workflows and other tests, but they cover the whole project, not specifically the skill's paths. There are no tests or evidence to validate this skill specifically. The doc is the primary source, but no external corroboration. Static review and lack of independent verification keep the score low.
- The skill relies on the 'vp' tool and commands that are not directly verified in the repo; test or confirm installation instructions (provided in README) before use.
- For mainland-China users, the services the T3 app depends on may not be directly accessible; the doc does not mention network reachability, so verify first.
- The skill enables sensitive operations (e.g., direct SQLite edits); though the doc advises caution and stopping the dev server before writes, ensure correct use of --base-dir to avoid touching shared data.
- There is no direct evidence linking this skill to the rest of the repo, so its actual effectiveness and maintenance status remain uncertain.
What does this skill do, and when should you use it?
This skill provides guidance for the web client of the T3 Code repository, enabling agents to start an isolated dev server, authenticate the browser on first navigation, and preserve the environment across turns. It covers pairing token handling, recovery of consumed or expired tokens, and direct SQLite inspection or seeding. It emphasizes environment lifecycle management, not tearing down until human review is complete. The skill is part of a larger repository that bundles four skills, and this one focuses specifically on the web experience.
It runs commands from the repository root, starting the full web stack with vp run dev, with optional --share for sharing across devices. It reads the dev runner output to determine server port, web port, base directory, and pairing URL. It manages browser authentication using one-time pairing URLs and token recovery commands. It inspects and seeds SQLite databases using node apps/server/scripts/t3-sqlite-state.ts, and uses the auth CLI for pairing tokens. It maintains the environment until the user confirms completion.
- A developer who needs to run the T3 Code web client locally and verify it loads correctly.
- A tester who wants to simulate first-time user authentication in a controlled browser using a one-time pairing URL.
- An agent that needs to inspect or reset test data while iterating on implementation.
- A developer who wants to keep the dev server running across multiple conversations without restarting.
- A QA engineer preparing an isolated test environment for another person without risking token consumption.
- An agent that needs to recover a consumed or expired pairing token to continue browser sessions.
What are this skill's strengths and limitations?
- Provides strict environment lifecycle management, preventing premature cleanup.
- Includes secure pairing authentication flow, protecting one-time tokens.
- Supports worktree-safe state isolation via isolated base directories.
- Offers troubleshooting guidance for common failures like token expiration and port shifts.
- Includes a reference for SQLite inspection and seeding for testing purposes.
- Relies on external tools `vp` and Node.js, which may require additional setup if not already installed.
- Only focuses on the web client, excluding mobile testing which requires another skill.
- The skill is in an early stage of the project and may be unstable, with no public docs site.
- No test suite is provided for the skill itself, and operations may change as the project evolves.
How do you install this skill?
The skill is located at .agents/skills/test-t3-app/SKILL.md in the repository. The full T3 Code repository is cloned or installed by the user (e.g., via npx t3@latest), and the skill folder should be placed in the standard Agent Skills location for your agent. Since it is part of the repo, there is no separate installation step. Dependencies include Node.js and the vp command-line tool.
How do you use this skill?
Trigger: Use this skill when an agent needs to run T3 locally, iteratively test UI behavior, recover from a pairing token issue, isolate dev state, or prepare test data in state.sqlite. Example prompt: 'Start the T3 web dev server and set up an isolated test environment.' Then follow the instructions: start the server with vp run dev, authenticate via the pairing URL, use t3-sqlite-state.ts for SQLite operations, and keep the environment alive until the task is complete.
How does this skill compare with similar options?
This skill is an in-repo skill for web testing of the T3 Code project, distinct from the test-t3-mobile skill for mobile clients. It is not a standalone third-party skill.