pinkrooster-terminal-ux
PinkRooster's terminal UI/UX design skill for coding agents.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/pinkroosterai--pinkroosterterminalux)HTML
<a href="/listings/pinkroosterai--pinkroosterterminalux"><img src="/api/listings/pinkroosterai--pinkroosterterminalux/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Choose your coding harness. Run shell commands in a terminal; start a fresh agent session once installation finishes.
| Harness | Recommended installation | |---|---| | Claude Code | `claude plugin marketplace add pinkroosterai/PinkRoosterTerminalUx`, then `claude plugin install pinkrooster-terminal-ux@pinkrooster-terminal-ux` | | Codex CLI | `codex plugin marketplace add pinkroosterai/PinkRoosterTerminalUx`, then `codex plugin add pinkrooster-terminal-ux@pinkrooster-terminal-ux` | | Antigravity CLI | `agy plugin install https://github.com/pinkroosterai/PinkRoosterTerminalUx` | | Pi | `pi install git:github.com/pinkroosterai/PinkRoosterTerminalUx` | | Cursor, Gemini CLI, GitHub Copilot, OpenCode | `npx skills add pinkroosterai/PinkRoosterTerminalUx`, then select your harness |
The skills CLI needs Node.js 22.20 or later; add `-g` for a user-wide install and `--copy` where symlinks are unavailable. Keep the whole `skills/terminal-ux/` directory, including `references/`. Install using one method per harness to avoid a duplicate skill.
**Verify before starting:** ask your agent, "Is the terminal-ux skill installed? Read it and summarise its two modes. Do not change any files." Expect a summary of build and audit mode. This checks discovery and loading, not behaviour; `docs/testing.md` has four fresh-session prompts that test behaviour.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 15 lines
- {
- "name": "pinkrooster-terminal-ux",
- "description": "PinkRooster's terminal UI/UX design skill for coding agents.",
- "owner": {
- "name": "pinkroosterai"
- },
- "plugins": [
- {
- "name": "pinkrooster-terminal-ux",
- "source": "./",
- "description": "Terminal UI/UX design skill: designs, builds and audits command-line output, interactive prompts and full-screen TUIs to a high UX standard, in any language or framework.",
- "version": "0.1.0"
- }
- ]
- }
README.md · README · 141 lines
- <div align="center">
- <img src="assets/logo.svg" alt="PinkRooster logo" width="160" />
- # PinkRooster Terminal UX
- **One portable agent skill that designs, builds and audits terminal interfaces (command-line output, interactive prompts and full-screen TUIs) to a high UX standard, in whatever language and framework your project already uses.**
- [](LICENSE)
- [](https://agentskills.io/specification)
- [](#development)
- [The idea](#the-idea) · [Install](#install) · [Usage](#usage) · [What's inside](#whats-inside) · [Compatibility](#compatibility) · [Development](#development)
- </div>
- ---
- ## The idea
- Coding agents write terminal UI that looks fine in a screenshot and breaks everywhere else:
- spinners leak into piped output, colour ignores `NO_COLOR`, shortcuts are hidden, layouts
- fall apart in an 80-column pane, Ctrl-C leaves the terminal broken, and errors say
- "Something went wrong". `terminal-ux` makes the agent apply the rules that prevent this
- while it writes the code, and audit existing interfaces against the same standard.
- The skill is **opinionated about behaviour and restrained about looks**. Stream discipline,
- graceful degradation (pipes, CI, no colour, narrow panes, no Unicode, screen readers), terminal
- restoration, discoverable keys and actionable microcopy are firm rules. Aesthetics are not
- imposed: the project's own look wins, and a project without one gets quiet, semantic defaults.
- It is stack-agnostic: it states what the user must see and be able to do, never how to use a
- particular UI library.
- These are instructions an agent follows, not a library or a linter. Your harness supplies file
- access, questions and shell commands. User instructions and project rules take precedence.
- ### When to Reach for It
- | If your immediate situation is... | Mode | What you get |
- |---|---|---|
- | A new command, script output or prompt flow to build | Build | Working code in your stack, plus the UX choices and the checks to run by hand |
- | A new full-screen screen, dashboard or picker | Build | A committed layout, keys with a legend, restoration on exit, a minimum-size message |
- | A small fix: table alignment, an error message, a flag's help text | Build | The edit, made directly from your code's context, without questions |
- | An existing tool whose terminal UX you want reviewed | Audit | Findings ordered by user harm, each with evidence and a fix; no code changed |
- It stays out of headless work (cron jobs, pure argument parsing, libraries with no
- human-facing output) and of web, desktop and mobile UI.
- ## Install
- Choose your coding harness. Run shell commands in a terminal; start a fresh agent session
- once installation finishes.
- | Harness | Recommended installation |
- |---|---|
- | Claude Code | `claude plugin marketplace add pinkroosterai/PinkRoosterTerminalUx`, then `claude plugin install pinkrooster-terminal-ux@pinkrooster-terminal-ux` |
- | Codex CLI | `codex plugin marketplace add pinkroosterai/PinkRoosterTerminalUx`, then `codex plugin add pinkrooster-terminal-ux@pinkrooster-terminal-ux` |
- | Antigravity CLI | `agy plugin install https://github.com/pinkroosterai/PinkRoosterTerminalUx` |
- | Pi | `pi install git:github.com/pinkroosterai/PinkRoosterTerminalUx` |
- | Cursor, Gemini CLI, GitHub Copilot, OpenCode | `npx skills add pinkroosterai/PinkRoosterTerminalUx`, then select your harness |
- The skills CLI needs Node.js 22.20 or later; add `-g` for a user-wide install and `--copy`
- where symlinks are unavailable. Keep the whole `skills/terminal-ux/` directory, including
- `references/`. Install using one method per harness to avoid a duplicate skill.
- **Verify before starting:** ask your agent, "Is the terminal-ux skill installed? Read it and
- summarise its two modes. Do not change any files." Expect a summary of build and audit mode.
- This checks discovery and loading, not behaviour; `docs/testing.md` has four fresh-session
- prompts that test behaviour.
- ## Usage
- | Harness | Explicit invocation example |
- |---|---|
- | Claude Code, plugin | `/pinkrooster-terminal-ux:terminal-ux audit the output of src/cli.py` |
- | Claude Code, standalone | `/terminal-ux audit the output of src/cli.py` |
- | Codex CLI / IDE | `$terminal-ux audit the output of src/cli.py` |
- | Pi | `/skill:terminal-ux audit the output of src/cli.py` |
- | Antigravity | Use its skill selector or ask "Use the terminal-ux skill to audit src/cli.py" |
- | Other skill-capable hosts | Ask for the skill by name; command syntax depends on the host |
- Natural requests such as "add a progress bar to the import command" or "review this TUI's
- keybindings" select the skill through its description. Selection is model-dependent;
- explicit invocation helps when routing is ambiguous.
- ## What's Inside
- A short [SKILL.md](skills/terminal-ux/SKILL.md) holds the workflow (classify the task, respect
- existing conventions, commit to a direction, build, audit), ten principles, the terminal-slop
- list and a routing table. The agent then reads only the references a task needs:
- | Reference | Covers |
- |---|---|
- | [output-and-resilience](skills/terminal-ux/references/output-and-resilience.md) | stdout vs stderr, TTY and CI detection, colour switches, Ctrl-C, terminal restoration, exit codes, progress and redraw |
- | [visual-style](skills/terminal-ux/references/visual-style.md) | colour roles, colour depth, light and dark backgrounds, attributes, glyph width, fallbacks, borders |
- | [layout-and-interaction](skills/terminal-ux/references/layout-and-interaction.md) | inline vs full-screen, TUI layout, size and resize, keys, legends, mouse, paging, tables, diffs |
- | [prompts-and-microcopy](skills/terminal-ux/references/prompts-and-microcopy.md) | prompts, validation, errors, success output, empty states, confirmations, help text |
- | [accessibility](skills/terminal-ux/references/accessibility.md) | screen readers, colour and contrast, motion, linear output, the accessible mode |
- | [audit-checklist](skills/terminal-ux/references/audit-checklist.md) | the pass/fail pre-delivery audit, terminal slop, the audit report format |
- Rules come from published guidance (clig.dev, `NO_COLOR`, terminal specifications, CLI
- writing guides and accessibility research); where the field is silent, a rule is labelled as
- the skill's opinion.
- ---
- ## Compatibility
- The common [Agent Skills format](https://agentskills.io/specification) makes the skill
- portable; it does not standardise tool APIs, permissions or command syntax.
- `argument-hint` is a deliberate Claude Code frontmatter extension. Strict `skills-ref`
- validation reports it as unsupported; this is documented rather than counted as a pass.
- Official references: [Claude Code](https://code.claude.com/docs/en/skills),
- [Codex](https://learn.chatgpt.com/docs/build-skills),
- [Antigravity](https://antigravity.google/docs/skills),
- [Pi](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/docs/skills.md),
- and [OpenCode](https://opencode.ai/docs/skills/).
- ## Development
- Contributor checks require Bash, standard Unix utilities and `jq`. Native package validation
- also requires the corresponding harness CLI; the skill itself has no runtime dependencies.
- ```bash
- bash scripts/check-skills.sh # skill, references, links, labels, packages
- bash scripts/test-check-skills.sh # plants faults in a scratch copy, expects rejection
- bash scripts/check-packages.sh # requires jq
- claude plugin validate .
- # Optional strict format validation (argument-hint is a known extension):
- npx -y skills-ref validate skills/terminal-ux
- ```
- Behaviour is tested by hand in fresh sessions with `docs/testing.md`. See
- [AGENTS.md](AGENTS.md) for maintenance rules: no UI library is ever named in the skill, the
- references stay between three and six files, and every one is reached from the routing table.
- Update the [changelog](CHANGELOG.md) and all package version records when preparing a release.
- The project is [MIT licensed](LICENSE). The header logo is shared with
- [PinkRooster](https://github.com/pinkroosterai/PinkRooster).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.