# Fast-lane norm — the single source

> **advisory**: nothing enforces the rules stated here. They are a judgement
> list, not a gate, and no lane enforces its own rules either — each lane only
> *states* its obligations: peaks-code's fast mode its step switches
> (`peaks code plan --fast`), peaks-race-code its completion proofs, which that
> skill itself calls self-reported and checked by nothing. So "stated here" does
> not mean "enforced somewhere".
>
> **The document itself is pinned by a test.** That is a different claim, and
> the next paragraph says exactly what the test checks and what it does not.

Two fast lanes share these two statements.
`tests/unit/standards/fast-lane-norm-single-source.test.ts` walks every markdown file under skills/, asserting the gate sentence appears once and the risk-list heading appears once. That is the whole guard — a copy of the risk list **under a different heading** is not caught, which is why this paragraph names the mechanism instead of promising more. The design record under `docs/` quotes the gate too, because a design doc has to name what it specifies; that is documentation drift, not lane drift. Two places that cannot be word-for-word identical should not have been split in the first place.

## 1. The acceptance gate

`test pass + tsc pass + lint pass` = GO.

A single QA round, **no** repair loop. peaks-code's fast mode turns the repair
loop off; peaks-race-code has no separate QA window to loop with.

## 2. When a task is not a fast-lane task

Any one hit means the task does not belong in a fast lane. **Each row is
annotated with how mechanical it actually is** — only row 6 is machine-checkable,
and the rest are judgement calls, which is worth knowing before treating this
list as if it decided anything on its own.

| # | Risk surface | Machine-checkable? |
|---|---|---|
| 1 | authn / authz / credentials / secrets handling paths | no — judgement |
| 2 | database schema or migrations | partly — the path can be matched (`migrations/` and the like) |
| 3 | public API surface (exported signatures change) | partly — codegraph / export-surface diffs help |
| 4 | concurrency / transaction semantics | no — judgement |
| 5 | dependency upgrades carrying API changes | partly — the `package.json` diff is mechanical, the API-change call is not |
| 6 | change size past a threshold | **yes — but no threshold exists yet** (spec U2), so this row is unactionable rather than decided. A lane has to set one before it can act on it |
| 7 | behaviour that cannot be believed from "run the tests once" | no — judgement |

**This list does not say what happens on a hit.** That belongs to each lane:
peaks-code's answer is the full workflow, and peaks-race-code's answer is
one-way escalation to peaks-code.