菜单
Choosing which Sage workflow fits a request. Use when the user asks what to run, which command or workflow to use, where to start, or when a request is ambiguous or spans multiple surfaces and needs to be routed to build, fix, architect, review, learn, reflect, or continue.
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。
When --quality-locked is active, loop review/revise at each Quality Gate until findings reach a clean bar (no Critical, no Major, only cosmetic Minor) or the iteration cap (10) is reached. Uses a deterministic Python checker for classification and decision logic; agent runs the actual review and revision steps.
Use after implementation passes the quality gates, when a change needs an independent pass over the code before it ships, or when the user asks to "QA this", "check the implementation", or "verify it matches the spec". Applies to Standard and Comprehensive scopes with the Task tool available; Lightweight tasks skip.
Use when the user picks [A] Review at a spec/plan/ADR/root-cause/fix-plan checkpoint, or asks for an independent review before implementation begins. Applies to Standard and Comprehensive scopes with the Task tool available; Lightweight tasks skip.
Use as the code-quality gate after spec-compliance and constitution checks pass, or when the user says "review my code", "quality check", "security review", or "is this code good". Runs as an independent sub-agent when the Task tool is available.
Adversarial review verifying implementation matches its specification — checks completeness (nothing missing) and precision (nothing extra). Distrusts the implementer's self-report. Use after implementation, or when the user says "check against spec", "does this match requirements", or "verify the implementation".
Configure Sage preset and project settings. Switch between base, startup, enterprise, or opensource constitution presets. Use when the user says "configure sage", "change preset", or "sage settings".
| name | sage-routing |
| description | Choosing which Sage workflow fits a request. Use when the user asks what to run, which command or workflow to use, where to start, or when a request is ambiguous or spans multiple surfaces and needs to be routed to build, fix, architect, review, learn, reflect, or continue. |
| version | 1.0.0 |
| type | system |
The eager layer carries the keyword map — the trigger. This carries the rest: what to do when the keyword map misses, how to present the choice, and what routing looks like when it is done well.
Layer 1 — Keyword routing (deterministic, checked first). The map in the eager layer. One match → confirm. Multiple matches → present them as options. No match → Layer 2.
Layer 2 — Sub-agent classifier. If the Task tool is available, spawn a classifier sub-agent:
Classify this request as UNDERSTAND (/research or /review --ux), ENVISION (/design or /architect), or DELIVER (/build or /fix).
Use the response to select the workflow → confirm. No Task tool → Layer 3.
Layer 3 — In-context classification (fallback).
| Shape of the request | Category | Workflow |
|---|---|---|
| Question, evaluation, "why" | UNDERSTAND | /research, /review --ux |
| Future, "should", "let's create" | ENVISION | /design, /architect |
| Action, "add", "implement" | DELIVER | /build, /fix |
| Ambiguous | — | present options from all matching categories |
Always, except for explicit slash commands and Tier 1 tasks:
Sage → [workflow]. [One-line rationale].
[1] [Workflow] — [skill → chain → arrows] ([N] steps)
[2] [Alternative] — [chain] ([N] steps)
[3] [Alternative] — [chain]
Pick 1-3, type / for commands, or describe what you need.
Show the chain and the step count. No time estimates — they are guesses wearing a number's clothing.
"audit our checkout UX" — keyword audit → /review --ux.
Sage → review workflow. Evaluating the checkout experience.
[1] Review — UX audit → evaluation (2 steps) [2] Research — interview → JTBD → opportunity map (3 steps) [3] Build — spec → plan → implement → verify
Pick 1-3, type / for commands, or describe what you need.
"the checkout page is throwing 500 errors" — keyword error → /fix.
Single match, no ambiguity, so confirm and go:
Sage → fix workflow. Investigating 500 errors on checkout.
"improve our onboarding conversion rate" — conversion → review,
improve → build. Two matches, so do not pick for the user:
Sage: This could go two ways:
[1] Review — UX audit → evaluation (2 steps) [2] Build — spec → plan → implement → verify [3] Research — interview → JTBD → opportunity map (3 steps)
Pick 1-3, type / for commands, or describe what you need.
"change the button color to blue" — Tier 1. Just change it. Routing a one-line change through a confirmation menu is the process becoming the point.
The routing chain exists because the alternative is an agent that treats every request as a DELIVER request — the default failure mode, and the one that ships a fix for a problem nobody diagnosed. If a request is a question, answering it with an implementation is not efficiency. It is a wrong answer that took work.