Repository navigation
Commit ce95164
docs(pm): name the landing flow's wait state when the ccr undraft route answers 429 (#19717)
Fixes #18743
## What lands
One rule line in
`.claude/skills/pm-dispatch/references/landing-operations.md` §B — the
landing
flow's own state list — naming what an arming run IS when the ccr
undraft route
(`POST .../pulls/{n}/ccr/ready_for_review`) answers 429. Before this, a
seat that met that 429
held a fact (the body's `resets at` instant), a prohibition (never retry
before the reset, never
busy-poll, never blind-retry) and **no declared outcome**, so its only
exits were to report a
dated wait as a blocked landing or to invent a cadence — the thing the
rules forbid.
Added at line 22, 120 bytes (the ratchet's per-line cap is 120, and a
line of exactly 120 bytes passes — 34 lines of the
corpus already sit at exactly 120):
```text
- undraft 撞 429 ⇒ 落地入等 `resets at` 态:该刻记卡,到点一次重试;⛔ 不报阻塞、不自定节奏。
```
It carries the three parts the card's contract asks for — the named wait
on the instant, the
instant recorded on the card, one retry at it — plus both prohibitions
(⛔ not a blocked landing,
⛔ no self-chosen cadence). The facts stay where they are:
`platform-readings.md` is not edited.
## Where it goes, and why there rather than the alternatives
Measured, not assumed. Three homes were on the table.
1. **Beside the arm act (chosen).** The 429 is returned **by the ready
flip itself** — zero bytes
written, `draft` unmoved. So the line belongs in the block that performs
the flip (the
flip-定点 → `绿即转 ready + 挂 auto-merge` → the no-busy-poll discipline). It
sits immediately
after the CI-webhook line so that the pair `绿即转 ready…` ↔ `CI success
webhook 不可靠…` stays
intact, and so the new state reads as the **permitted action** under
that line's ⛔ 不坐等 /
⛔ 不忙轮询: the wait is a 定点 the platform names, not one the seat picks.
2. **Beside the three-step enqueue check — rejected.** That check is
gated on
`ready + 全绿 ≠ 已入队`. A 429 leaves the PR *not ready at all*, so that
precondition is false
and the three-step ladder never starts. Filing the state there would put
it behind a door the
429 can never open.
3. **`rest-channel.md` — rejected.** That file is a per-operation
**channel** table ("which route
does this act take"), and it already carries the route. What is missing
is not a route but a
**landing state**, and the landing flow's state list is §B here.
## How the line is paid: the retired duplicate
`landing-operations.md` is at ceiling 69 with headroom 0, so the line is
paid in-file by retiring
one. Retired:
```text
- ⛔ 不放宽其它任何 PR 的 queue-only 落地。
```
Two independent covers, both checkable:
- **It restates a rule that lives in `SKILL.md`.** `SKILL.md` already
states
`路径面干净的才转 ready → 入队;队列是唯一被认可的落地路径,⛔ 永不队列外合并。` — unconditional, and
strictly stronger than "don't relax it for any other PR". This file's
own ceiling convention,
quoted in the ratchet's `CEILINGS` comment for the 80 → 69 lowering, is
verbatim
**"no rule already stated in SKILL.md"**. So the line is a duplicate by
the file's declared standard,
not by my reading of it.
- **Inside this file it has no residual work.** Its only possible job
was bounding the two
main-red exceptions directly above it, and both already carry their own
bounds in their own
text — `限 p0、机械、根因已核实的止血 PR` and `仅限 main-red 修复`. Nothing normative is
lost with
it gone.
Net: 69 lines before, 69 after. The ceiling is **not** raised and no
cross-file move is declared.
## Verification
Derived with `node scripts/pm/dispatch-gates.mjs --repo
objectstack-ai/objectstack --commands`
(no paths — the script takes its own change set from the merge base),
then reconciled with
`--ran`. Exit codes captured before any pipe, into a file.
```text
dispatch-gates --ran: 17 derived famil(ies) accounted for — 17 run, 0 NOT-MEASURED
(a DERIVED zero — all 17 recorded an exit code and none of them is 3).
```
The families the dispatch named, with the verdict each printed:
| gate | exit | verdict line |
|:---|:---|:---|
| `pnpm check:pm-skill-ratchet` | 0 | `landing-operations.md is 69 lines
(ceiling 69; headroom 0)` · `widest table row is 0 bytes (pin 0;
headroom 0)` |
| `pnpm check:pm-skill-id-lint` | 0 | `30 file(s) clean (pattern
/#[0-9]{3,}/g)` |
| `pnpm check:skill-frame-sync` | 0 | `the one declared copy of the
decision frame is internally coherent` · 77 md files scanned for
undeclared copies |
| `pnpm check:pm-governed-prose` | 0 | `2 instruction surface(s) name
all 6 registered governed surfaces and claim no others` |
| `pnpm check:doc-authoring` | 0 | `403 files clean — no bare metadata
literals` |
| `pnpm check:nul-bytes` | 0 | `scanned 9210 text file(s) … no raw ASCII
control bytes` |
The other eleven derived families also exited 0:
`check:pm-governed-merges`,
`check:cross-package-test-inputs`, `check:driver-memory-census`,
`check:agent-test-spelling`,
`check:gitlink-declared`, `check:refd-timer-probe`,
`check:watch-hint-literal`,
`@objectstack/lint run check:doc-formula-expressions`,
`check-closing-keyword-parity.mjs` (both
with and without `--self-test`), `check-comment-mask-corpus.mjs` and
`check-harness-current.mjs --self-test`. Four of them first answered
**exit 3** in the fresh
worktree (prerequisite not met — no `node_modules`; one also needed
`@objectstack/formula` + `@objectstack/lint` built). Exit 3 is not a
finding and was not recorded
as one: `pnpm install` and a locked
`turbo run build --filter=@objectstack/formula
--filter=@objectstack/lint` cleared it and all four
were re-run to a real verdict. The build ran through
`scripts/pm/os-verify-lock.sh`, which printed
`VERDICT command-exit 0 · held the lock 177s · waited 0s`.
Beyond the gate: a direct control-character sweep of the edited file,
`grep -naP '[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]'`, exited 1 (no match).
No ablation or reverse-verification run is claimed here: the change is
one line of governed prose
with no executable behaviour, so there is nothing whose removal a test
could be shown to detect.
The ratchet is the instrument that would have caught a miscount, and it
is green at 69/69.
## Changeset
`skip-changeset`. Measured rather than asserted: `.claude` exists as
exactly **one** directory in
the tree, at the repo root. The only package at the repo root is
`@objectstack/spec-monorepo`,
which is `private: true` and declares no `files[]`. Of the 83 tracked
`package.json` files, 70
declare a `files[]` and **zero** of their entries names `.claude` under
any spelling; the
remaining 13 are all `private: true`. So nothing in this diff reaches a
published tarball.
## Acceptance notes
No out-of-scope cards are owed from this round. One boundary worth
recording, deliberately **not**
filed:
- `SKILL.md` and `references/filing-gate.md` each already carry a 429
rule (`429 即停写` /
`收到 429 或次级限流 ⇒ 该席本 fire 停写`). Those govern the **write budget** on the
claude[bot]
token, a different pool from the session credential's GraphQL hourly
pool that the undraft
route draws on, and they prescribe *stopping writes* — which is
precisely what would push a
seat toward reporting the landing as blocked. They are near neighbours
of the gap this PR
closes, not statements of it, and they are correct as written. An
observation, so no card.
---
_Generated by [Claude
Code](https://claude.ai/code/session_01Wnstp2kTth7sGXfr8fXypc)_
Co-authored-by: Claude <noreply@anthropic.com>1 parent e2c55ed commit ce95164
1 file changed
Lines changed: 1 addition & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
19 | 19 | | |
20 | 20 | | |
21 | 21 | | |
| 22 | + | |
22 | 23 | | |
23 | 24 | | |
24 | 25 | | |
| |||
47 | 48 | | |
48 | 49 | | |
49 | 50 | | |
50 | | - | |
51 | 51 | | |
52 | 52 | | |
53 | 53 | | |
| |||
0 commit comments