st4ck-lite

st4ck-lite

Record tests with your AI agent. Replay them forever, for free.

The free Claude Code plugin for st4ck. Captures agent-driven sessions as deterministic Playwright test files stored as plain markdown in your repo. No account, no key, no service.

# Inside Claude Code:
/st4ck-lite:author https://app.example.com "Sign in as alice and verify the dashboard loads"
# Agent walks the site, runner captures every primitive, writes ./tests/sign-in-as-alice.md

# Replay anytime — zero LLM, zero cost:
/st4ck-lite:run tests/sign-in-as-alice.md

That's the whole product. Two slash commands. Two npm packages underneath (st4ck brand binary + the runner it wraps). Zero magic.

---

Three beats

1. Agent ergonomics. A focused IPC vocabulary — actions (navigate, click, fill, press, select, check_box, hover, upload, wait_until, evaluate), text-based disambiguation (click_by_text, hover_by_text, type_by_text), conditional dispatch (branch), observation (snapshot, url), control (continue, abort) — over line-delimited JSON on stdin/stdout. Per-call opt-in flags (dispatch_chain, dispatch_events, atomic) handle no-code platform runtimes (Bubble, Retool, Webflow, n8n, Wix Velo, Glide, FlutterFlow). Your agent already drives a browser; this is the smallest surface that lets it.

2. Authoring-by-use. Your agent isn't writing a test — it's using the site. The recording is the test. Walk through the flow once; you get a markdown file you can rerun, version, share.

3. Deterministic free replay. Recorded md files have zero LLM calls. Reruns are pure Playwright, 10–15× faster than the recording, $0 marginal cost. Run them in CI, on a cron, or whenever you push.

---

Free vs paid

Capability	st4ck-lite (this plugin)	st4ck paid platform
10-primitive IPC vocab	✓	✓
Agent-driven recording	✓	✓
Deterministic markdown replay	✓	✓
Locator-priority ladder (Tier-1 self-heal)	✓	✓
Auto-wait + strict locators (Playwright-grade)	✓	✓
Block-format teaching methodology	✓	✓
LLM self-heal on selector drift (Tier-2)	—	✓
Cross-project knowledge base	—	✓
TRIAD attestation + server-enforced review	—	✓
13-point review checklist with cross-validation	—	✓
Agent Teams authoring orchestration	—	✓
PRD/spec/dev-task binding + impact analysis	—	✓
Coverage reporting against intent sources	—	✓
Security test generation pipeline (4-phase)	—	✓
Multi-project + multi-environment orchestration	—	✓

→ Full platform: st4ck.io

---

What lives in this repo

• .claude-plugin/plugin.json — Claude Code marketplace metadata
• st4ck/skills/ — record + replay skills (the open subset of methodology)
• st4ck/commands/ — slash command aliases (/st4ck-lite:author, /st4ck-lite:browse, /st4ck-lite:run)
• st4ck/agents/ — recording sub-agent (uses local md files; no MCP key)

The runner itself is the st4ck-runner npm package — same binary the paid plugin uses. The lite plugin doesn't fork the runner; it constrains the SKILL set to what works without an app.st4ck.io connection.

---

Install

In Claude Code:

/plugin marketplace add edo-ceder/st4ck-lite
/plugin install st4ck-lite@st4ck-lite-marketplace

Then /reload-plugins to activate.

Or use the standalone CLI without a plugin (@latest always resolves to the current release):

npx st4ck@latest author https://example.com "Sign in as alice and verify the dashboard loads"
npx st4ck@latest browse launch https://example.com --record --out ./tests/sign-in-as-alice.md
npx st4ck@latest run ./tests/sign-in-as-alice.md

The npm package is st4ck (binary name st4ck). For CI / reproducibility, pin to a specific version (e.g. npx [email protected] …); see npm view st4ck versions for the list. The plugin manifest schema has no version-pinning field, so pinning happens at invocation time.

---

Status

0.1.0-alpha.0. Public alpha; the CLI surface + md format are stable. Skill content + recording UX iterates through 0.1.x based on alpha feedback.

License: Apache-2.0.