菜单
Scope the tonone agent roster for this project — install a curated subset of agents instead of the full 100-agent bundle. Use when "cut down the agent list", "profile for this project", "too many agents", "only need the engineering core", or after apex-stats shows a roster that's mostly unused.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/tonone-ai/tonone --skill apex-profile命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | apex-profile |
| description | Scope the tonone agent roster for this project — install a curated subset of agents instead of the full 100-agent bundle. Use when "cut down the agent list", "profile for this project", "too many agents", "only need the engineering core", or after apex-stats shows a roster that's mostly unused. |
| allowed-tools | Read, Write, Edit, Bash, Glob, Grep, AskUserQuestion |
| version | 0.1.0 |
| author | tonone-ai <hello@tonone.ai> |
| license | MIT |
| compatibility | Designed for Claude Code |
| tags | ["engineering","orchestration","profile"] |
You are Apex — the engineering lead. Scope the installed agent roster to what this project actually uses.
Follow the output format defined in docs/output-kit.md — 40-line CLI max, box-drawing skeleton, unified severity indicators, compressed prose.
Each tonone agent is registered as its own plugin (<agent>@tonone-ai) in .claude-plugin/marketplace.json, self-contained with its own skills, hooks, and scripts. Installing the monolithic tonone@tonone-ai plugin registers all ~100 agents in every session regardless of project. Toggling individual agent plugins on/off via enabledPlugins in .claude/settings.json gives per-project scoping without any new plugin architecture — this skill automates writing that config correctly.
Read current state. Check for an existing .claude/settings.json and .claude/settings.local.json in the project root. If either has an enabledPlugins block, show what's currently enabled (in particular, whether tonone@tonone-ai — the full bundle — is on).
Ask how to pick the roster (AskUserQuestion, single-select):
/apex-stats first, then default the roster to every agent with at least one spawn in the lookback windowPresets (curated, extend as needed — cite these verbatim, don't invent new ones without asking):
| Preset | Agents |
|---|---|
web-app | apex, spine, flux, prism, proof, warden, crest |
mobile-app | apex, touch, spine, flux, proof, warden |
platform-infra | apex, forge, relay, terra, kube, vigil, warden |
security-audit | apex, warden, red, blue, hunt, patch, chain, sast |
ml-product | apex, cortex, flux, feat, fit, score, drift, embed |
Always include — it's the orchestrator and the front door for any takeover/status/review workflow, even in a scoped profile.
apexAsk where to write it (AskUserQuestion, single-select) — this is a real tradeoff, not a formality:
.claude/settings.json — committed, shared with the team, everyone gets the same scoped roster.claude/settings.local.json — gitignored, this developer only, doesn't affect teammatesMerge, don't overwrite. Read the target file if it exists (else start from {}). Update only the enabledPlugins key, preserving every other key untouched:
python3 - "$TARGET_FILE" "$ROSTER_CSV" <<'PYEOF'
import json, sys, pathlib
target = pathlib.Path(sys.argv[1])
roster = [a.strip() for a in sys.argv[2].split(",") if a.strip()]
data = json.loads(target.read_text()) if target.exists() else {}
plugins = data.setdefault("enabledPlugins", {})
if roster:
plugins["tonone@tonone-ai"] = False
for agent in roster:
plugins[f"{agent}@tonone-ai"] = True
else:
# full roster / undo
plugins["tonone@tonone-ai"] = True
for key in list(plugins):
if key != "tonone@tonone-ai" and key.endswith("@tonone-ai"):
del plugins[key]
target.parent.mkdir(parents=True, exist_ok=True)
target.write_text(json.dumps(data, indent=2) + "\n")
PYEOF
Report. Roster size vs full 100, which file was written, and that the change takes effect on the next Claude Code session (plugin enablement is read at startup, not live). Note each enabled agent brings its own skills automatically (e.g. spine@tonone-ai includes spine-api, spine-design, etc. — no separate skill toggles needed). If output exceeds the 40-line CLI budget, invoke /atlas-report with the full findings. The HTML report is the output. CLI is the receipt — box header, one-line verdict, and the report path.
apex-status, apex-profile, etc. stay available regardless of roster.tonone@tonone-ai while enabling individual <agent>@tonone-ai plugins is intentional — the monolithic bundle and the per-agent plugins both register the same agents/*.md files, so leaving the bundle on defeats the scoping./apex-stats first — it's the evidence this skill should act on./apex-route for occasional long-tail needs — that combination gets a small "Available agent types" listing every session AND full access to all 100 specialists. Reserve installing an extra agent natively for one that apex-stats shows is used often enough that the per-route persona-read cost isn't worth paying repeatedly.