AGENT SKILL

ce-setup

Check RocketClaw's tools and repo config and fix what it finds, scaffold a Compound Pack with `pack:<id>`, or draft and refresh coding standards and review docs from pull request history. Use when setting up or diagnosing RocketClaw, or when creating or updating CODING_STANDARDS.md or REVIEW.md, including on a schedule.

License 许可
MIT
最近更新
2026-11-14
Stars
★ 0
Forks
0

01 前言 Front Matter

name
ce-setup
description
Check RocketClaw's tools and repo config and fix what it finds, scaffold a Compound Pack with `pack:<id>`, or draft and refresh coding standards and review docs from pull request history. Use when setting up or diagnosing RocketClaw, or when creating or updating CODING_STANDARDS.md or REVIEW.md, including on a schedule.
argument-hint
[pack:<id> | standards] [mode:non-interactive]

02 资源文件 19

03 SKILL.md 正文

RocketClaw Setup

OpenCode V2 installation prerequisite

Document this action for the user; do not edit host/global configuration or a consuming project's OpenCode files. Ask the user to merge "experimental": { "subagent_depth": 3 } into ~/.config/opencode/opencode.json. Use opencode debug paths or opencode debug paths config to locate the actual global configuration directory when different, and update its existing opencode.json or opencode.jsonc; never create a competing JSON/JSONC file or replace the entire file. Preserve unrelated settings, sibling experimental fields, and any existing integer depth higher than 3. Deep agentic delegation requires depth at least 3. Installing the plugin or shipping package-local opencode.json does not merge that configuration into consumers, and higher-precedence configuration must not lower the effective depth below 3.

Configured peer/model choices, tiers, and effort remain authoritative intent. Execute routes using native opencode.models discovery and native subagents/shell, not another harness dispatcher or acpx. Preserve compatibility examples and provider references as configuration vocabulary, not executable cross-harness routes. Skill-requested delegation satisfies a host user-or-skill permission rule, but never an unconditional prohibition. Delegation permission, model-override permission, and remaining nesting depth are separate: inherit the model when overrides are forbidden, classify actual dispatch errors, and report unavailable required model independence or coverage as incomplete rather than claiming success. Never bypass a denial or depth limit through another harness.

Run JJ from the target workspace's absolute root. For repository-scoped GitHub CLI commands use (cd "$workspace_root" && GIT_DIR=$(jj git root) gh ...). Store all temporary data under that root's ignored .tmp/. JJ reference: https://docs.jj-vcs.dev/latest/cli-reference/ .

Non-interactive runs

Run non-interactively when the arguments contain mode:non-interactive, or when the request makes unattended intent unmistakable, such as a scheduled or CI job, or a caller asking for a run "headless", "unattended", or "without questions". "Automatically" alone is not that signal: it says when setup runs, not whether anyone can answer. When the signal is absent or unclear, run interactively. Strip the token before reading the rest of the arguments. A non-interactive run asks nothing and follows references/non-interactive.md, which replaces every offer, question, and approval in this file and its references.

Interaction Method

In an interactive run, ask each question below with the host's blocking question tool, and never silently skip or auto-configure. Read references/communication.md from this skill's directory before your first offer. It says how to reach that tool and what each offer, question, and preview must tell the user.

Pack Scaffold

When the invocation names a Compound Pack to add, create, or scaffold (the pack:<id> argument, or the same request in words), read references/pack-scaffold.md from this skill's directory and follow it in place of Phases 1-2: it writes the pack and its config entry only after the user approves, runs the health check itself, and reports into Phase 3 (Summary).

Review Docs

When the invocation asks to draft, refresh, or update coding standards or review docs (the standards argument, or the same request in words), read references/project-docs.md and follow it in place of Phases 1-2, drafting the whole set of files even when the request names one. Report into Phase 3.

Artifact Root Resolution

Every RocketClaw skill that reads or writes an artifact tree (solutions, plans, and others) resolves its root by this rule. Step 2 reports the resolved root.

<!-- ce-docs-root:start --> Resolve the artifact root <root> before composing any artifact path.

  • Read docs_root from <repo-root>/.rocketclaw/config.yaml only (<repo-root> = jj workspace root). Do not read it from config.local.yaml. Unset -> <root> is docs, exactly as before.
  • Validate a set value: a repo-relative directory whose real, symlink-resolved path stays inside the repo and is neither the repo root nor under .git/. Otherwise stop with an error naming docs_root and the value -- never fall back to docs.
  • Use <root> as the sole artifact location: create it if absent, compose each path as <root>/<subdir> with this skill's own subdirectory, and never also read docs. <!-- ce-docs-root:end -->

Phase 1: Diagnose

Step 1: Determine Plugin Version

Detect the installed RocketClaw plugin version by reading the plugin metadata or manifest when the platform exposes it. If the version cannot be determined, skip this step.

If a version is found, pass it to the check script via --version. Otherwise omit the flag.

Step 2: Run the Health Check

Before running the script, display:

RocketClaw -- checking your environment...

Run the bundled check script. Set SKILL_DIR to the absolute directory you loaded this ce-setup SKILL.md from — the Bash tool's CWD is the user's project, not the skill dir, so a bare scripts/ path will not resolve:

SKILL_DIR="<absolute path of the directory containing this SKILL.md>";
if [ -f "$SKILL_DIR/scripts/check-health" ]; then bash "$SKILL_DIR/scripts/check-health" --version VERSION; else echo "Bundled health script not found at $SKILL_DIR/scripts/check-health; run the inline checks from ce-setup instead."; fi

Use the same command without --version VERSION if Step 1 could not determine a version.

If the script is unavailable, run the inline equivalent listed in references/repo-fixes.md.

Display the diagnostic output to the user. The health report includes the resolved artifact root and which config layer supplied it (per Artifact Root Resolution above); show that line so the operator can confirm where artifacts will be written. Missing config.yaml is a reported absence, not a project issue.

Step 3: Decide Whether Fixes Are Needed

Repo-local fixes the health report names apply only to the checkout that report diagnosed. If Phase 2 will write to a different writable checkout, diagnose that checkout first. Session-level findings such as plugin version and optional tools still come from this session's Phase 1.

After the health report, decide Phase 2 from writable-checkout availability:

  • If this session has a writable JJ workspace, run Phase 2 locally, including when project_issues is 0. Phase 2 always refreshes the example and always offers to create config.yaml when that file is missing.
  • If this session has no writable checkout, but the user named a repository and the harness exposes a remote repo-work surface with a writable checkout, run Phase 2 on that surface instead and report the remote repo-local fixes in Phase 3.
  • Otherwise skip Phase 2 and go to Phase 3, saying repo-local writes were skipped because no writable checkout is available.

Two offers concern this machine rather than the checkout, so they stand whether or not Phase 2 runs. If the report names a legacy Compound Codex tool map, offer to remove it following references/legacy-codex-tool-map.md from this skill's directory. Optional compatibility package-cache maintenance is described in references/acpx-cache-warm.md; it is not the active OpenCode peer execution path.

Phase 2 remediates every project issue the report names, each in its own step of references/repo-fixes.md. That includes an out-of-date config.yaml, retired config keys, a ce-work implementation engine the report marks unavailable or invalid, and malformed dormant work_engine_preferences or work_engine_effort.

Phase 2: Fix Repo-Local Issues

Read references/repo-fixes.md from this skill's directory before making any repo-local change. It remediates every project issue the report names and carries Steps 4-10: removing the obsolete rocketclaw.local.md, refreshing the example config, refreshing out-of-date config files with their settings kept, offering to create config.yaml, repairing invalid work_engine_preferences and docs_root, the two .gitignore offers, and the agent-instructions offers (a knowledge-store mention, the compounding directive, and the chat-register directive for ce-noslop), and the offer to draft the project's review docs.

Every change to a user-owned file is offered and applied only if the user approves.

Phase 3: Summary

User-runnable invocation rendering. In setup summaries, default to /ce-setup and /ce-help; use $ce-setup and $ce-help only when the active host is Codex or explicitly documents dollar-prefixed skill invocation. On oh-my-pi (omp), keep /ce-setup and use /skill:ce-help. Render only the invocation as inline code and output one form only.

Write the summary that references/summary.md describes. End it with these two lines, with the rendered invocations filled in:

Run `<rendered invocation>` anytime to re-check.
Run `<rendered ce-help invocation>` to ask which skill fits or what comes next.

04 同仓库更多 Skills