From tiktest
Scaffolds tiktest.md config file by inferring dev-server URL, start command, and auth flow from README, package.json, framework configs. Use for tik-test setup in new projects or missing config.
How this skill is triggered — by the user, by Claude, or both
Slash command
/tiktest:setupThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
The user wants a real, persistent `tiktest.md` in their project so they can run `/tiktest:run` / `/tiktest:quick` without re-deriving the URL, start command, and login each time. Inspect the project, infer what you can, ask the user for what you can't, and write the file.
The user wants a real, persistent tiktest.md in their project so they can run /tiktest:run / /tiktest:quick without re-deriving the URL, start command, and login each time. Inspect the project, infer what you can, ask the user for what you can't, and write the file.
Bail if a config already exists. Check the cwd for either filename:
ls tiktest.md tik-test.md 2>/dev/null
If either prints, stop and report: "Found existing config at <path>. Edit it directly, or delete it first if you want to regenerate." Do not overwrite.
Inspect the project — opportunistically, not exhaustively. Read whatever config and source files give the strongest signal about how the project runs locally and whether it has auth. Pick what's relevant per project; do not enumerate every possible file. The Read tool handles missing files gracefully — try the obvious candidates and stop once you have enough signal.
Useful sources (sample, don't exhaust):
package.json, pyproject.toml, Gemfile, Cargo.toml, go.mod) for project name + description + dev-server scripts.README.md — look for sections like "Local dev", "Getting started", "Development", "Install", "Login", "Demo credentials".vite.config.*, next.config.*, astro.config.*, svelte.config.*, vue.config.*, nuxt.config.*, webpack.config.*, vercel.json, turbo.json).app.py or manage.py, Sinatra config.ru, Go main.go for http.ListenAndServe).Makefile / justfile for dev / start / run targets..env.example for default ports and any committed dev credentials.docker-compose.yml for service ports.If the answer's already in package.json and README.md, you're done — don't read the rest of the source tree.
Synthesize a draft tiktest.md from what you found. The shape is:
# <Project name>
<One-paragraph description of what the project does and the primary surfaces.>
## Login
<Auth instructions — only include this section if you found evidence of an auth flow.>
## Local dev
start: <command that starts the dev server>
The project serves at <URL>.
Fill what you can confidently infer:
name + description, or the README's H1 + first paragraph.http://localhost:5173, a Next.js default → http://localhost:3000) or an explicit port from a config file. Use http://localhost:<port>.npm run dev, pnpm dev, yarn dev, python3 -m http.server <port>, cargo run, etc)..env.example, a "## Demo credentials" section in the README, a hard-coded test user in source). Do NOT invent credentials.Identify gaps and ask one question at a time. Anywhere you didn't find clear evidence, ask the user. Wait for the answer before asking the next question. Common gaps:
http://localhost:3000)"npm run dev)"Ask each question separately. Do not bundle them. If the user gives you everything in one reply, that's fine — incorporate and move on.
Show the user the draft before writing. Render the markdown in chat (inside a fenced block) and ask:
"Here's the draft
tiktest.mdI'm about to write:<draft>Want me to write it as-is, or tweak anything first?"
If the user wants edits, incorporate the feedback and re-render. Loop until the user approves.
Write and confirm. Use the Write tool (not a Bash heredoc — same shell-injection rationale as the other skills) to write the approved markdown to <cwd>/tiktest.md. Then tell the user:
"Created
tiktest.md. You can edit it directly any time, or run/tiktest:runor/tiktest:quickto start using it."
tiktest.md / tik-test.md. Bail per Step 1.Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub marcushyett/tik-test --plugin tiktest