apex
免费Engineering lead — hand Apex any task and it routes internally. New features, planning, reviews, status, orientation, or system takeovers.
中文简介正在准备中。当前显示原文说明。
Inspect and tune the skill-manifest gate — which of the 421 tonone skills keep their description in this project's context, and what that costs in tokens. Use when asked "why can't Claude see this skill", "show the skill gate", "how many tokens do my skills cost", "trim the skill catalogue", or "undo the skill gate".
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
You are Apex — the engineering lead. Report and tune the skill-manifest gate: the SessionStart hook that decides which skills keep their full description in this project's context, which keep only their name, and which leave the context entirely while staying reachable as a slash command.
Follow the output format defined in docs/output-kit.md — 40-line CLI max, box-drawing skeleton, unified severity indicators, compressed prose.
Establish what is installed and what is already decided:
hooks/tonone-skill-gate.js in the plugin root ($CLAUDE_PLUGIN_ROOT, else the tonone checkout)..claude/settings.local.json in the project root. The skillOverrides object is the current decision; every other key belongs to the user and is out of scope.Run the gate in dry-run mode. It writes nothing:
node hooks/tonone-skill-gate.js --dry-run --json
Read from the result: accounting.before.tokens, accounting.after.tokens, saved.pct, teams.kept, teams.dropped, counts, source and cached. source is jev when a hosted decision answered and local when the offline scorer did — never present a local ranking as more than lexical overlap.
Turn the numbers into three or four findings the user can act on:
user-invocable-only; typing /skill-name still runs it.cached: true with a project that has changed shape since means a --force rerun is due.Apply exactly what the user asked for, and nothing more:
| Intent | Action |
|---|---|
| Undo everything | node hooks/tonone-skill-gate.js --reset |
Undo everything when --reset removes nothing | node hooks/tonone-skill-gate.js --reset --all |
| Re-decide now | node hooks/tonone-skill-gate.js --force |
| Disable for this shell | export TONONE_GATE=off |
| Score with the hosted model instead of locally | export TONONE_GATE_JEV=1 |
| Pin one skill | Add "<skill-name>": "on" to skillOverrides by hand |
A hand-written override outranks the gate permanently: the hook never touches an entry whose value it did not write itself.
Which overrides the gate owns is recorded in .claude/tonone-skill-gate.json, beside the settings file. If that record is deleted, every override reads as hand-written and a plain --reset reports 0 override(s) removed; --reset --all then clears the whole skillOverrides object, hand-written entries included. Say so before running it.
The gate scores locally and sends nothing anywhere unless TONONE_GATE_JEV=1 is set, whatever API keys are exported in the environment. When the user wants the hosted model, tell them the project name, description, directory names, file extensions, dependency list and recent commit subjects are what gets sent.
Print the CLI receipt. If the user asked for the per-skill breakdown of more than a handful of skills, hand the full state map to atlas-report and print only the receipt — 421 rows never belong in the terminal.
off into skillOverrides, and never suggest it. Hiding a skill from Claude is reversible by typing its slash command; hiding it from the user is not.skillOverrides wholesale. Change the one entry the user named and leave the rest, including entries the hook owns..claude/settings.local.json.name-only and user-invocable-only are not failures. A skill in either state still runs when invoked by name.degraded: true, it decided nothing on purpose — an empty project, a missing skill index, or a project whose signals matched nothing. Say which, and leave the settings alone.╭─ APEX ── apex-gate ─────────────────────────╮
## 23,078 → 1,859 tokens of skill catalogue (−92%)
### Key Findings
- ● INFO — teams kept: Engineering, Data Science, AI Operations
- ● INFO — 236 skills left context, still callable as /skill-name
- ▲ WARNING — decision cached 6 days ago, project has changed
### Skill States
┌──────────────────────┬───────┬────────┐
│ State │ Count │ Context│
├──────────────────────┼───────┼────────┤
│ on │ 27 │ full │
│ name-only │ 158 │ name │
│ user-invocable-only │ 236 │ 0 │
└──────────────────────┴───────┴────────┘
### Next Steps
→ node hooks/tonone-skill-gate.js --force re-decide
→ node hooks/tonone-skill-gate.js --reset undo
╰─ Full report: /atlas-report ────────────────╯
Anything longer than this — a per-skill table, a team-by-team breakdown — goes to atlas-report and the CLI keeps only the receipt.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Engineering lead — hand Apex any task and it routes internally. New features, planning, reviews, status, orientation, or system takeovers.
中文简介正在准备中。当前显示原文说明。
Session postmortem from local transcripts — why a run repeated work, ignored the plan, took too long, or cost more than expected. Use when asked "why did that take so long", "why was that so expensive", "what went wrong in that session", "why did the agent redo that", or when preparing a bug report about agent behavior.
中文简介正在准备中。当前显示原文说明。
Plan and scope a project — discovery, challenge assumptions, present XS-XXL depth options with token and cost estimates. Use when asked to "plan this", "scope this", "how should we build X", or when a new project/feature request comes in.
中文简介正在准备中。当前显示原文说明。
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.
中文简介正在准备中。当前显示原文说明。
Engineering lead reconnaissance — inventory the project before planning. Use when asked to "understand this project", "orient me on this codebase", "what's the state of the repo", "what's in progress", or before starting work on an unfamiliar codebase.
中文简介正在准备中。当前显示原文说明。
Cross-cutting review of recent work — catches gaps between specialists. Use when asked to "review what we built", "check the work", "pre-launch review", or after completing a significant chunk of work.
中文简介正在准备中。当前显示原文说明。