원클릭으로
setup-statusline
Live token counter in your CLI. Shows real-time input/output/cache token counts in the Claude Code status bar
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Live token counter in your CLI. Shows real-time input/output/cache token counts in the Claude Code status bar
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Plan-then-execute sprint orchestrator. Works like a real delivery team: split work into sprints (~1 human-week each), plan them all up front WITH the user, then autonomously run each sprint through a full cycle (research→plan→design→do→QA→fix→deploy) to completion. Multiple sprints can run at once (concurrent dispatch). The leader (main session) dynamically scaffolds project-local agents for whatever domain — not dev-only (marketing, research, ops, data all fit). bkit-aware: borrows bkit agents/skills internally when present, runs fully standalone otherwise. Not for single-file edits, one-shot bug fixes, or work under ~a few hours.
Invoke for "/cowork-insights" command or when the user asks to summarize, review, or report on past Claude Code sessions. Analyzes sessions to show key prompts (verbatim), structured assessments (goal/outcome/friction), tool usage patterns, and actionable insights. Produces HTML report + shareable Markdown for Jira/Notion/Slack. Three report formats — full (deep narrative), standard (core insights), minimal (quick team share). Supports --from/--to with absolute (2026-03-01) or relative (7d, 2w, 1m) dates. Trigger on phrases like weekly status update, sprint recap, what did I do with Claude, AI usage patterns, session history, minimal recap, what I worked on today, share with team, cowork-insights. DO NOT invoke for active tasks (debugging, refactoring, code review, project setup) or for commit-time recaps (use cowork-commit instead).
Trigger whenever the user asks to commit AND wants the commit message enriched with AI collaboration history. Creates a lightweight commit message (key decision highlights + link) and a full directive-log file with conversation transcript + recap. The key signal is the combination of (1) making a commit with (2) capturing how AI contributed. Trigger on phrases like commit with AI recap, attach collaboration history to commit, record AI work in commit, cowork-commit. DO NOT trigger for plain commits without AI documentation, standalone time-period recaps (use cowork-insights instead), PR reviews, or general git operations.
One-time bootstrap of an existing project's docs/ and source into the cowork-doc-sync taxonomy structure. After a detailed gap analysis, relocate docs to match the standard. Phase 1 = relocation only (no new creation, includes moving content between docs), Phase 2 = analyze source to create new docs (only after user approval). For ongoing maintenance use /cowork-doc-sync. Triggers: cowork-doc-init, /cowork-doc-init, init doc structure, relocate docs, organize existing docs, doc init, doc bootstrap
Ongoing doc-sync skill that aligns a project's docs/ with the current code/decision state. Call once at the very end, after implementation/refactoring is complete. Enforces a numbered taxonomy (00-reference~99-misc) + status model (LIVING/ACTIVE/FROZEN) + migration rules. To fit an existing project into this structure for the first time, use /cowork-doc-init. Triggers: cowork-doc-sync, /cowork-doc-sync, sync docs, align docs, organize docs, doc sync, doc alignment
Compare DevMD files against actual source code. Measures coverage, accuracy, and consistency with deterministic counting and evidence-backed findings.
| name | setup-statusline |
| description | Live token counter in your CLI. Shows real-time input/output/cache token counts in the Claude Code status bar |
| when_to_use | Use when user wants to install, uninstall, or configure the token-saver statusline. Triggers on "setup statusline", "install statusline", "statusline install". |
Manage the claude-code-token-saver statusline in Claude Code's ~/.claude/settings.json.
ONLY show help if the user's argument literally contains the word "help" (e.g. /setup-statusline help). If no argument or any other argument is given, SKIP this section entirely and proceed to Step 1.
If the user provides "help" as argument, show usage summary and stop:
/setup-statusline — Install or uninstall the live usage monitor
Format:
[RUN🟢] $0.10/$12.23 | [5H🟢] 9% ⏳1h32m | [W🟡] 65% ⏳1d3h | [CTX🟢] 22%
───── last / total 5h window usage weekly (≥60% only) context window
(subscribers only)
Options:
install Register statusline (default)
uninstall Remove and restore previous config
help Show this help
Examples:
/setup-statusline install
/setup-statusline uninstall
Do not modify any settings. Just display the help text and stop.
install (default if no arg) — register the statusline scriptuninstall — remove and restore backup if availableRead ~/.claude/settings.json
Check current statusLine field:
Case A: No statusLine — install directly.
Case B: Already our script — check if path matches current ${CLAUDE_PLUGIN_ROOT}.
If version differs, update path silently, then print "Updated to current version." and stop.
If same version, print "Already installed." and stop.
Check by matching the command path containing claude-code-token-saver/scripts/statusline-logger.sh.
Case C: Different statusLine exists — warn and ask the user:
⚠️ Existing statusline detected:
command: {existing command}
Options:
1. Replace (backup existing to _statusLineBackup)
2. Cancel
Which? (1/2)
Wait for user response. If "1", backup to _statusLineBackup then install. If "2", stop.
Install: set statusLine to:
{
"type": "command",
"command": "${CLAUDE_PLUGIN_ROOT}/scripts/statusline-logger.sh"
}
Verify the script is executable:
chmod +x "${CLAUDE_PLUGIN_ROOT}/scripts/statusline-logger.sh"
Output:
✅ Statusline installed.
It will appear after your next interaction with Claude.
Subscriber format:
[RUN🟢] $0.10/$12.23 | [5H🟢] 9% ⏳1h32m | [W🟡] 65% ⏳1d3h | [CTX🟢] 22%
API key format:
[RUN🟢] $0.10/$12.23 | [CTX🟢] 22%
Fields:
- [RUN] $0.10/$12.23 — **Current user turn cost** / cumulative cost since `claude` started.
The turn cost accumulates across follow-up tool calls so a $1.43 warning
stays visible until you start a new prompt. A new turn is detected when
no API call happens for 60 seconds (tunable via
`CC_TOKEN_SAVER_TURN_IDLE_SEC`). Resets when you exit and restart claude.
- [5H] 9% ⏳1h32m — Anthropic 5-hour rate limit usage (subscribers only).
⏳ shows time until the window resets.
- [W] 65% ⏳1d3h — 7-day weekly rate limit. Only shown when ≥60%.
⏳ shows time until the weekly window resets.
- [CTX] 22% — Context window usage. Higher = more tokens per call.
Color thresholds:
| Indicator | 🟢 Normal | 🟡 Warning | 🔴 Critical |
|-----------|-----------|------------|-------------|
| RUN | < $0.30 | ≥ $0.30 | ≥ $1.00 |
| 5H | < 70% | ≥ 70% | ≥ 90% |
| W | (hidden) | ≥ 60% | ≥ 90% |
| CTX | < 35% | ≥ 35% | ≥ 70% |
When 5H reaches 🔴, → /report-limit hint appears.
Other warnings show → /usage-view current for detailed analysis.
To remove: /setup-statusline uninstall
Read ~/.claude/settings.json
Check current statusLine:
Not our script — print "claude-code-token-saver statusline is not installed." and stop.
Our script — remove statusLine field. If _statusLineBackup exists, restore it to statusLine and delete _statusLineBackup.
Output:
✅ Statusline removed.
{If restored: "Previous statusline restored from backup."}
{If no backup: "No previous statusline to restore."}
node -e for JSON manipulation (safe parsing, no jq dependency).statusLine and _statusLineBackup.${CLAUDE_PLUGIN_ROOT} to stay portable.