메뉴
Use when you want to check if Claude Code is installed and ready, or when troubleshooting the plugin
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
Use when you want to delegate a coding task to Claude Code — it executes the task and returns results for review
Use when you want to check the status of a Claude Code task — see progress, phase, results, or list all jobs
Use when you want to cancel a running Claude Code task
Use when you want to review code changes made by Claude Code — standard review for bugs, or adversarial review to challenge implementation choices
| name | setup |
| description | Use when you want to check if Claude Code is installed and ready, or when troubleshooting the plugin |
Verify that Claude Code is installed and the plugin is ready to use. Performs static checks (zero model calls): CLI protocol verification, source-vs-cache comparison, model routing classifier status, and state schema health. An optional cost-bearing liveness probe can be enabled with explicit authorization and a budget guard.
Call cc_setup with cwd set to the absolute path of the user's current workspace.
If everything is ready, inform the user they can start delegating tasks.
If issues are found:
npm install -g @anthropic-ai/claude-code--print --input-format text --output-format json). Suggest updating Claude Code to 2.1.208+.claude auth or setting ANTHROPIC_API_KEY--print, --input-format, and --output-format flags are supported (print-mode JSON capability)model selector. The plugin does not read, write, or modify any external routing configuration.match, differ, or not-installed when running from source). Never prints a green compatibility claim without an actual comparison.When the user explicitly asks for a real Provider liveness check, call cc_setup with:
livenessProbe: truetimeoutSeconds: <positive integer> (required — wall-clock budget for the probe)maxBudgetUsd: <positive number> (required — cost ceiling for the probe)model (optional): same selector semantics as cc_delegate (inherited / alias / native)This makes one real model call and incurs a cost. It is NOT a free check. The probe runs a trivial task ("Reply with exactly: OK") and reports the result, cost, and duration. It is blocked if any static check fails.
Honest cost + private evidence: When telemetry is missing the probe reports cost as unknown (never $0.00). A private, bounded, auditable artifact is persisted under the probe ID recording the route snapshot, route status, execution/model evidence, duration, exit/failure classification, and cost with explicit provenance. A usage key is never treated as execution proof — only transcript evidence is. The MCP output links only to the probe ID and a safe summary.
Budget guard (fail-closed): Before the probe spawns Claude Code, the setup verifies that the installed CLI supports the --max-budget-usd flag (inspected via claude --help). If the flag is unsupported, the probe fails closed — no Provider call is made. The probe also fails closed if livenessProbe is not literally true, if timeoutSeconds is missing or non-positive, or if maxBudgetUsd is missing or non-positive.
Do not enable the liveness probe without explicit user authorization. Do not enable it in CI or automated setup flows.
cc_setup with the current workspace's absolute cwdcc_setup with the current workspace's absolute cwdcc_setup with cwd, livenessProbe: true, timeoutSeconds: 30, maxBudgetUsd: 0.25cc_setup with cwd, livenessProbe: true, timeoutSeconds: 30, maxBudgetUsd: 0.25, model: "Opus"