# Live Console

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-punkaze-skills-live-console`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [punkaze](https://agentstack.voostack.com/s/punkaze)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [punkaze](https://github.com/punkaze)
- **Source:** https://github.com/punkaze/skills/tree/main/plugin/skills/live-console

## Install

```sh
agentstack add skill-punkaze-skills-live-console
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# live-console

Spin up a **console**: a zero-dependency local web app, rendered from a
declarative **spec**, that persists every response the user gives to a store on
disk and self-terminates on Submit — so you read their answers back without
copy-paste. The engine (`scripts/console_server.py`) owns all page behavior;
you only author the spec.

## Run loop

1. **Choose the surface.** Start from a **template** — list `templates/` and
   read each file's `_about` line for its purpose — or write a spec from
   scratch. Copy your starting point into the session scratchpad and fill in
   real items.
   *Done when:* a spec file exists in the scratchpad with a `title` and ≥1 item,
   every field has a `type` + `key`, and any referenced images sit beside the
   spec. Anything the schema can't express → raw-HTML escape hatch (see
   `AUTHORING.md`).
2. **Launch (background).** Run, as a background process:
   `python3 "${CLAUDE_PLUGIN_ROOT}/skills/live-console/scripts/console_server.py" `
   *Done when:* the printed `url` opened in the browser and you've noted the
   printed `results` file path.
3. **Hand off and wait.** Tell the user the console is open and to click
   **Submit & done** when finished. Do NOT poll. *Done when:* the background
   process **exits** (Submit) — that exit is your wake signal — or the user says
   to check now.
4. **Read and report — exhaustive, visually grouped.** Read the results file.
   Report as a compact table or status-grouped list (✅/❌/⏭) rather than a
   paragraph per item. Account for **every item in the spec**: state each
   item's response, and explicitly flag any left unanswered or marked
   fail/reject, drilling into fails. *Done when:* no spec item is silently
   omitted — each is named as answered (with its value) or flagged
   unanswered/failed.
5. **Clean up.** If the server is still running (the user read early without
   submitting), stop it. *Done when:* no console process running, its port free.

## Field types (detail in `AUTHORING.md`)

`toggle` (pass/fail/skip, approve/reject) · `comment` · `choice` (single/multi) ·
`rating` (stars/number) · `rank` (drag-reorder) · item `visuals` (compare HTML
snippets or image files → pick one + comment each; a single entry doubles as a
plain illustration — a diagram beats a paragraph, see AUTHORING.md).

**Visual-first:** if an item's content is inherently spatial (a flow, a layout,
a before/after, an architecture) — show it, don't describe it. See
AUTHORING.md's "Visual-first authoring."

## Facts

- **Store:** `.results.json` beside the spec — `{_meta:{done,submittedAt}, responses:{itemId:{key:value,updatedAt}}}`. Auto-saved on every change, so it's readable mid-session; `_meta.done` flips to `true` only on Submit.
- **Notify:** the server runs in the background and **exits on Submit** — no polling; the exit is the signal to read the store.
- **Defaults:** binds `127.0.0.1` only, OS-assigned free port, auto-opens the browser (macOS), theme-aware, ephemeral scratchpad store. Flags: `--port N`, `--no-open`, `--results PATH`.
- **Schema + raw-HTML escape hatch (`Console.set/get/submit`):** see [`AUTHORING.md`](AUTHORING.md).

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [punkaze](https://github.com/punkaze)
- **Source:** [punkaze/skills](https://github.com/punkaze/skills)
- **License:** Apache-2.0

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-punkaze-skills-live-console
- Seller: https://agentstack.voostack.com/s/punkaze
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.