What does this skill do, and when should you use it?
This skill lets Claude operate the full configuration surface of claude-statusbar (cs) on your behalf. You say things like "switch to dracula", "the warning color is too pale", or "the bar isn't showing", and Claude maps the intent to the right cs command and runs it. It replaces the older /statusbar family of slash commands and explicitly guards against wrong upgrade paths (cs upgrade is the only supported one). It fits Claude Code users who already run or plan to run claude-statusbar.
Translates user intent into specific cs subcommands executed via Bash: inspect/switch theme, style, and density; preview all 27 combinations; override the three severity colors (ok/warn/hot) independently; run cs doctor self-checks (settings. statusLine block, binary on PATH, daemon liveness, stale caches); upgrade via cs upgrade, which detects the actual install channel; toggle roughly 40 settings including session cost, cache countdown, todo progress, active tool, subagents, session duration, lines changed, git ahead/behind, and the forecast chip; auto-detect no-quota mode for third-party relays/Bedrock/Vertex (context window promoted to its own battery bar); and manage the AgentParty/Codex bridge line and daemon fast mode.
- Heavy Claude Code users who want 5h/7d rate-limit headroom and reset countdowns at a glance, without switching to a separate window.
- Users on third-party relays (ANTHROPIC_BASE_URL != official) or Bedrock/Vertex, where no official quota exists and the context-window battery replaces it.
- Users who want a theme by vibe but forget names: say "something soft and warm, dark" — Claude suggests twilight and confirms before switching.
- Users unhappy with one color but not the whole theme: override color_warn and friends directly.
- Anyone whose bar disappears or burns CPU: cs doctor self-checks, or recommends daemon fast mode when refreshInterval ≤ 2s (CPU drops from ~6% to ~2% at 1Hz).
- AgentParty/Codex users who want a local-only line showing channel, identity, and unread counts in the bar.
How do you install this skill?
- Daily auto-upgrade is enabled by default (auto_upgrade true) and updates every install channel in the background; disable with `cs config set auto_upgrade false` or CLAUDE_STATUSBAR_NO_UPDATE=1 if unwanted.
- The curl|bash one-liner writes to ~/.claude/settings. and may register a macOS login item (HUD); security-conscious users should read install.sh first.
- The optional per-model weekly limits feature reads ~/.claude/.credentials. or the macOS Keychain and queries an undocumented Anthropic endpoint; it is off by default — do not enable unless needed.
- This score is a static source review only; rendering and installation were not executed, and publisher identity is unverified by FollowSkills.
- Fast mode (daemon) runs persistently with periodic refresh; review docs/daemon.md for CPU and cache-file behavior before enabling.
- Shell / CLI
- Network access
- Local filesystem
cs CLI (claude-statusbar, pip/uv/binary install)Claude Code statusLine integration (cs --setup)
Install the underlying tool first: the one-liner curl -fsSL https://raw.githubusercontent.com/leeguooooo/claude-code-usage-bar/main/install.sh | bash (prebuilt binary, no Python needed), or pip install claude-statusbar then cs --setup — which wires the statusLine hook and installs the skill. Restart Claude Code. The skill lives at skills/claude-statusbar/ in the repo. Steps for installing the skill file alone (without cs) are not fully documented; see the skill-only path in docs/install.md.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/leeguooooo/claude-code-usage-bar.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/claude-statusbar" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- switch theme to nord
- turn off the cache countdown
- the bar isn't showing
- make the safe color sharper
After setup, trigger it conversationally: "switch theme to nord", "turn off the cache countdown", "the bar isn't showing", "make the safe color sharper", "upgrade to the latest". The skill's decision tree matches the right cs command and returns a one-line confirmation. Note: upgrade only via cs upgrade — never let Claude guess pip install -U; destructive commands (cs config reset, cs daemon stop) require your confirmation first.
What are this skill's strengths and limitations?
- Single conversational entry point covering cs's entire config surface, replacing six legacy slash commands.
- Explicit guardrails: no hand-editing config files, no guessed upgrade commands, confirmation before destructive actions.
- Dedicated layout and auto-detection for no-quota (relay/Bedrock/Vertex) setups.
- The doctor flow covers the four most common failure modes, including the real gotcha that a new theme only applies on the next render.
- Hard dependency on cs being installed and statusLine wired; the skill itself installs nothing.
- Some features are experimental (bar_shimmer) or off by default (subagent display, tool rollup).
- Older macOS onefile releases leaked a ~11 MB _MEI* temp dir per killed refresh; v3.32.5+ is required.
- The optional per-model weekly limit reads OAuth credentials and hits an undocumented Anthropic endpoint — off by default but privacy-sensitive.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Claude Status Bar Control Skill this page | 62 · Recommended | ★ 377 | 16d ago | MIT |
| Azure AI Gateway Governance ✓ Microsoft · Official | 55 · Use with care | ★ 2.7k | 3mo ago | MIT |
| EAS Update Health Insights | 55 · Use with care | ★ 2.9k | 20d ago | MIT |
| Jetson Headless Memory Optimizer ✓ NVIDIA · Official | 55 · Use with care | ★ 3.6k | 11d ago | Apache-2.0 |
| ClawSec Advisory Feed | 47 · Use with care | ★ 1.1k | 12d ago | AGPL-3.0 |
The README positions cs as an embedded status line, complementary to standalone TUIs: ccusage (long-form usage analytics, cost breakdowns) and Claude Code Usage Monitor (burn-rate forecasting). cs aims for one line and one decision per second; the TUIs suit users wanting charts and aggregates in a side terminal, and the two coexist.
How did FollowSkills review this skill?
The skill is a thin wrapper over the local `cs` CLI invoked via Bash; it explicitly requires user confirmation before destructive commands (`config reset`, `daemon stop`), discloses data flows well (local caches; the AgentParty bridge 'never reads tokens or makes network requests'; per-model weekly limits' read-only OAuth access is opt-in and documented), and a SECURITY.md defines the threat model and reporting path. Deductions: daily auto-upgrade is on by default (a background update channel, though off-switches exist), the curl|bash install pattern, and unverified publisher identity.
The decision tree covers install/config/diagnose paths; `cs doctor` gives structured self-checks (settings., PATH, daemon, stale caches) with concrete diagnostic flows; the repo has CI (3.9–3.12 matrix) and committed tests covering key backend paths (account switching, projection stores). Static review cannot execute; tests cover the backend rather than the skill invocation path itself, and failure-feedback quality on abnormal input is only partially confirmable — capped at 10 per static calibration.
Trigger words are extremely thorough (bilingual Chinese/English, synonyms like cs/状态栏/theme/threshold), with explicit non-fit boundaries (no-quota mode, narrower Codex bridge), input/output examples and a 27-combination preview; Chinese-language support is good. Deduction: some behavior depends on specific Claude Code internals (statusLine, transcript heuristic), and boundary statements are scattered across docs rather than centrally declared.
Layered docs (SKILL decision tree → segments/styles/configuration/troubleshooting guides), MIT license, version 3.42.0 in pyproject, CHANGELOG and Releases links, CONTRIBUTING and maintainer contact, a single sanctioned upgrade path (`cs upgrade`) with rationale. Deductions: SKILL.md itself does not state its version mapping, many defaults are implicit in tables, and the legacy slash-command deprecation has no timeline.
Clear value proposition with a comparison table (vs ccusage, claude-monitor); output is an immediately visible one-line status bar and commands are single-intent and directly usable; the claim of official 5h/7d data (not local estimates) is a strong signal. Static review cannot verify rendered output correctness, and comparative benefit rests on author claims — conservatively 6 within the static cap.
Multiple auditable materials exist: committed tests including a real incident post-mortem (test_account_switch), CI workflows with release smoke tests and benchmark scripts, and SECURITY/CHANGELOG governance. Static review cannot independently reproduce conclusions and not all claimed 900+ tests were reviewed — capped at 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →