메뉴
Use when an agent needs to call the `claude` CLI from another runtime, delegate work to Claude Code non-interactively, resume a Claude session, capture Claude output, or coordinate cross-runtime agent handoffs.
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 finishing a feature branch and wanting a final-state cleanup pass. Audits changed comments, structures, and defensive fallbacks with parallel subagents, fixes every validated finding caused or exposed by the change, and leaves unrelated baseline debt untouched.
Use when the user asks for a combined code-review and cleanup final pass before integration, including per-commit review-and-cleanup. Runs both passes to convergence without expanding beyond issues caused or exposed by the original change.
Capture intentional changes made directly to chezmoi targets back into the dotfiles source — modifier templates and directly managed files alike. Use when application-managed or manually edited dotfiles should be synchronized into the dotfiles repository for review without applying changes to the live home directory.
Use when a Pull Request is ready for final pre-squash verification, description and title synchronization, squash-message drafting, and an optional user-authorized squash merge.
Use when reading Claude Code .jsonl session transcripts — covers locating files, filtering rewinds and compactions, focused skimming, and subagent transcripts.
Use when reading Codex .jsonl session transcripts — covers locating files (incl. by UUID or project), filtering rewinds and compactions, focused skimming, and subagent transcripts.
| name | using-claude-cli |
| description | Use when an agent needs to call the `claude` CLI from another runtime, delegate work to Claude Code non-interactively, resume a Claude session, capture Claude output, or coordinate cross-runtime agent handoffs. |
Use command claude -p for non-interactive delegation. Keep model, tools, settings, and environment-derived defaults unchanged unless the user explicitly asks otherwise.
Always invoke the binary via command claude, never bare claude. A shell alias or function named claude (common in interactive profiles) would otherwise intercept the call and inject its own flags or wrapper behaviour; command bypasses that and runs the real executable.
log=$(mktemp)
session_id=$(node -e 'console.log(crypto.randomUUID())')
command claude -p --output-format json \
--session-id "$session_id" \
--disallowedTools EnterPlanMode,AskUserQuestion \
--dangerously-skip-permissions \
"$PROMPT" </dev/null | tee "$log" >/dev/null
jq -e '.is_error == false' "$log" >/dev/null \
|| { echo "claude: failure (see $log)" >&2; jq -r '.result // .' "$log" >&2; exit 1; }
result=$(jq -r '.result' "$log")
Pass $session_id to resume for follow-up turns.
</dev/null is required. Without it claude -p waits 3s for stdin (visible warning "no stdin data received in 3s") before proceeding — a flat 3s/call penalty when invoked from any non-TTY caller (subagents, CI, Bash tools).--session-id <uuid> pre-assigns the session id, so parallel callers do not have to read the result before they can resume. node -e 'console.log(crypto.randomUUID())' works on macOS and Linux; node ships with claude itself.jq -e '.is_error == false' assertion is required. claude exits 0 on every failure path tested: unknown CLI flag (prints error: plain text, no JSON), missing prompt, API 404, permission denial. Trusting $? will silently treat error messages as the answer. The same assertion also fails when stdout is not JSON, covering CLI usage errors.thinking/tool_use, user tool_result, etc.) use --output-format stream-json --verbose instead; the same --session-id and </dev/null rules apply.Resume when the next prompt depends on the previous Claude context. Always resume by the captured $session_id. Do not use -c — it picks the most recent conversation in the current directory and will clobber parallel callers in the same cwd.
log=$(mktemp)
command claude -p --output-format json \
-r "$SESSION_ID" \
--disallowedTools EnterPlanMode,AskUserQuestion \
--dangerously-skip-permissions \
"$PROMPT" </dev/null | tee "$log" >/dev/null
jq -e '.is_error == false' "$log" >/dev/null \
|| { echo "claude: failure (see $log)" >&2; jq -r '.result // .' "$log" >&2; exit 1; }
result=$(jq -r '.result' "$log")
--model, --tools, or other config overrides unless the task specifically requires them.claude unless a human is actually driving the TUI.--output-format json for automation because it exposes session_id, result text, errors, and usage.