Skip to content
Agentmart
Submit a tool

verify-agentlings

use this when verifying Agentlings live — the desk, the world, the inbox — against a running local install

Skill by briant92

IndexedApache-2.0
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/briant92-verify-agentlings

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

use this when verifying Agentlings live — the desk, the world, the inbox — against a running local install

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/verify-agentlings

What's inside

Skill name
verify-agentlings
Description
use this when verifying Agentlings live — the desk, the world, the inbox — against a running local install
License
Apache-2.0
Instructions
82 lines in SKILL.md · view on GitHub

Agent Skills spec issues

  • files: referenced but missing: scripts/control-agentlings.mjs

Files (5)

  • SKILL.md
  • README.md · features/
  • desk.md · features/
  • inbox.md · features/
  • world.md · features/

SKILL.md (first lines)

# Verify Agentlings

Closed-loop verification for the Agentlings app: doctor → drive → screenshot →
evidence. Prefer the control CLI over throwaway click scripts.

**Repo:** `briant92/Agentlings`
**Where it runs:** this machine, loopback. API/WS on `http://127.0.0.1:4600`,
web dev server on `http://127.0.0.1:5173`. After `npm run build` the API port
serves the web bundle too (one origin, D-272). Override with `AGENTLINGS_BASE_URL`.
**CLI:** `node scripts/control-agentlings.mjs`
**Feature maps:** `features/`
**Owner of the prove:** Nest. **PROVE_STATUS:** not yet.
**Firewall:** Agentlings only — not FO/IGPL. Never put `AGENTLINGS_PASSWORD`,
API keys, `.env` contents or session cookies into maps, evidence or captions.

## Launch

1. Prefer a loopback install: `npm install`, copy `.env.example` → `.env` (Brian
   fills it; do not invent credentials), then `npm run serve` (stable, no file
   watching — D-140). Use `npm run dev` only while changing server code.
2. One-origin: `npm run build`, then the server alone on `:4600`. This is what a
   container does; `dev`/`serve` do not need it.
3. On loopback with no `AGENTLINGS_PASSWORD` there is no gate (D-271). Anywhere
   else the gate is on and the app opens on the login screen. This skill does
   not hold the password: stop at login evidence and mark `inconclusive`.
4. A hosted install refuses supervised live acting and anything else that needs
   the operator's desktop or disk (README: *What an install cannot do hosted*).
   Do not drive those there; prefer loopback or a Nest-owned install.

## Doctor

```bash
node scripts/control-agentlings.mjs doctor
# optional: --dry-run  (prints the planned probes only)
```

Probes `GET /api/session` — the one ungated route — on each candidate origin
and reports `{ required, authed }`. No server answering → `ok: false` with an
`inconclusive` line and exit code 1. **Do not claim green.**

## Drive

Read the Feature Map for the surface first, then the CLI:

```bash
node scripts/control-agentlings.mjs open-desk
node scripts/control-agentlings.mjs open-inbox
node scripts/control-agentlings.mjs screenshot --out .agentlings/evidence/title.png
```

All commands print **JSON** on stdout. The app is a single-origin SPA with no
deep links — every screen is the same URL — so `open-*` returns the URL plus
the click path and the DOM selectors, not a route. `screenshot` captures what
the app shows on a cold open (login or title screen); reaching a level is the
Phase-2 drive step and is not built yet.

## Evidence

- Save screenshots and JSON under `.agentlings/evidence/` (gitignored) or a
  PR-attached path. Never under `features/`.
- A claim of done requires an artifact path **or** an explicit
  `inconclusive: …`. Matches PROJECT.md's definition of done: show the evidence.
- `npm test` (vitest) is complementary; it does not replace a live doctor →
  drive → screenshot. The repo's `scripts/prove-*-ui.mjs` scripts are the
  existing live probes and read `.env` themselves — this CLI never does.

## Cleanup

- Each CLI command closes the browser it opened.
- Do not leave headed browser windows on the shared desktop; supervised live
  acting is headed by construction and is not this skill's business.
- Do not commit `.env`, `browser-state*.json`, session cookies or evidence.
- `node scripts/control-agentlings.mjs cleanup` — best-effort note; no
  persistent control browser exists yet.

## Maintain

Audit Feature Map drift against `web/src/` (screens, panels, world) and
`DECISIONS.md`. PR only inside this skill directory and
`scripts/control-agentlings.mjs`; never "fix" a product bug by editing a map.

From briant92/Agentlings, Apache-2.0.

More from briant92/Agentlings

Signals

Stars
0
Forks
0
Open issues
6
Last commit (this folder)
Oct 15, 2026 (13 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from briant92

Cite this page

verify-agentlings skill. Agent Mart. https://agentmart.dev/skills/briant92-verify-agentlings (last updated 2026-10-24)