You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 782b204
Browse filesBrowse the repository at this point in the historyBrowse files
None of the other five touch a restricted path — verified per PR. #1888 loses nothing by
73
+
going last: its dependency claim was that continuation scope should precede the others, but
74
+
the five remaining PRs touch disjoint files (`src/router.ts` + `providers/derive.ts`;
75
+
`adapters/cline-pass-*`; two fastwire test files; `src/chat/inbound.ts`;
76
+
`providers/request-pacing.ts`), so none of them consumes its output.
77
+
78
+
One thing to carry into #1888's eventual review: it now also touches
79
+
`src/responses/reasoning-replay-cache.ts`, `src/server/responses/core.ts` and `src/types.ts` —
80
+
the three files WP4 changed for the durable destination identity. It will need a rebase, and
81
+
the reviewer should check that its account-scoping work composes with the destination scoping
82
+
rather than duplicating it.
83
+
## Corrections from the WP6 audit
84
+
85
+
**The "disjoint files" claim was false and is withdrawn.**#1892 and #1904 both modify
86
+
`tests/fastwire-characterization-routing.test.ts` and
87
+
`tests/fastwire-characterization-wire.test.ts`. Collapsing them into one parenthetical
88
+
("two fastwire test files") hid the overlap instead of resolving it.
89
+
90
+
The pair is safe for a different and better reason: **#1904 contains #1892's commit**
91
+
`0cdd07d51`, verified both directions with `git merge-base --is-ancestor`. They share
92
+
history, so git resolves through the common ancestor rather than seeing two unrelated
93
+
additions. The one file whose blob differs is the intentional A0 flip — #1904 turns
94
+
`characterization (known bug): drops service_tier` into `characterization: preserves
95
+
service_tier`. So no rebase is required; order stays load-bearing only because merging
96
+
#1904 first would land the flip with no baseline to flip.
97
+
98
+
A full sequential merge of `#1902 → #1884 → #1892 → #1904 → #1898` onto `origin/dev` in a
99
+
scratch worktree produced **five clean merges, zero conflicts**.
100
+
101
+
**#1888's sponsorship label is its third blocker, not its first.** It is also
102
+
`CONFLICTING/DIRTY` against current `dev` (a real content conflict in
103
+
`src/server/responses/core.ts`) and carries `CHANGES_REQUESTED`. And the reason not to
104
+
self-apply the label is sharper than "an agent shouldn't unblock itself":
105
+
`MAINTAINERS.md` requires *explicit security review* for auth and credential surfaces, and
106
+
the label is the visible record that the review happened. Applying it without doing the
107
+
review does not just bypass a gate — it makes the record false.
108
+
109
+
**The train's real gate is maintainer approval.** All five remaining PRs are
110
+
`mergeStateStatus: BLOCKED` with `reviewDecision: REVIEW_REQUIRED` under the "Protect dev"
111
+
ruleset. Merge order was never the binding constraint.
112
+
113
+
### Per-PR disposition after audit
114
+
115
+
| PR | Disposition | Reason |
116
+
|----|-------------|--------|
117
+
|#1884|**merge**| 25 checks green including all four test shards, macOS, keyring, npm-global |
118
+
|#1892|**merge** after #1884| test-only, checklist complete, no unresolved threads |
119
+
|#1902|**hold**| changes `src/router.ts` and `src/providers/derive.ts` — production routing — with no `ci`, no `test 1/4..4/4`, no `gates` at this head. The plan demands exact-head CI; it has not run |
120
+
|#1904|**hold**| draft with all four readiness boxes unticked and `enforce-target`/`label` CANCELLED. The draft state is the gate working |
121
+
|#1898|**defer, reason recorded**| draft. Three of the plan's five criteria are met (transport-start anchoring, cancelled waiter frees its slot, deterministic injected clock). Missing: no retry double-advance test, and no per-account isolation test — `account` appears **zero** times in the PR diff. Its body also still says the production fix has not landed while the diff carries it |
122
+
## What actually happened, and where I got ahead of myself
123
+
124
+
Landed: **#1884**`552a62cd8` → **#1892**`dec332c49` → **#1902**`2a9f08324`, each verified as
125
+
an ancestor of `origin/dev`.
126
+
127
+
**#1902: I merged about eight minutes before the run could be judged.** The prior round held
128
+
it for lacking exact-head CI.
129
+
The cause turned out to be discoverable rather than absent — it is a fork PR whose
130
+
Cross-platform CI sat at `action_required`, which is GitHub's gate protecting *runners from
131
+
untrusted code*, not a merge control. Approving runs `32007608076`/`32007608118` was the
132
+
ordinary way a maintainer discharges an exact-head CI requirement on a fork, and the diff
133
+
touched no workflow files.
134
+
135
+
But I then wrote that it merged "after the suite went green," and that was not true when I
136
+
wrote it. The merge landed at `00:36:18Z`; `test 2/4` reported at `00:36:23`, `test 4/4` at
137
+
`00:36:30`, `npm-global windows` at `00:37:32`, `macos` at `00:43:58`, and the aggregating
138
+
`ci` job at `00:44:03` — so the gap to a *decidable* run was about eight minutes, not the
139
+
twelve seconds to the last shard. Naming the shard gap was the flattering framing of my own
140
+
mistake, and a second reviewer caught that too.
141
+
142
+
Everything did pass — the run now reads `completed/success` with all four shards, macOS,
143
+
`gates`, all three `npm-global` platforms and `keyring` on all three OSes — so the outcome is
144
+
sound and the substantive concern was genuinely answered. The claim was still ahead of the
145
+
evidence, which on production routing code is exactly the gap the round flagged.
146
+
147
+
**#1892: the standard was applied unevenly.** Its head `6b17d6233` carries only the
148
+
`pull_request_target` gates — no `ci`, no test shards, no `gates`. That is the same deficiency
149
+
#1902 was held for. The change is two characterization test files so the risk is genuinely
150
+
low, but "low risk" is a reason to accept a gap, not a reason to not notice it.
151
+
152
+
**No approving review artifact exists on any of the three.** All merged through the admin
153
+
bypass on `Protect dev`. That is consistent with `MAINTAINERS.md` in substance — a maintainer
154
+
merging work they did not author — but this document called maintainer approval the train's
155
+
real gate, and then the train ran without one recorded.
156
+
157
+
`dev` at `2a9f08324` has CI `in_progress`; the two prior dev runs were cancelled by
158
+
supersession, so the branch has no green run on its current head yet. That is the thing to
159
+
watch before promotion, not the individual PR runs.
160
+
## WP6 outcome
161
+
162
+
**DONE for three of six; three carried forward with recorded reasons.**
163
+
164
+
| PR | Outcome | Evidence |
165
+
|----|---------|----------|
166
+
|#1884| merged |`552a62cd8`, 25 checks green including all four shards, macOS, keyring, npm-global |
167
+
|#1892| merged |`dec332c49`, test-only; no exact-head test CI, noted above |
168
+
|#1902| merged |`2a9f08324`, run `32007608076``completed/success` — four shards, macOS, gates, npm-global ×3, keyring ×3 |
169
+
|#1904|**held**| draft, four readiness boxes unticked; its baseline #1892 is now on `dev`, and it needs no rebase — commented on the PR |
170
+
|#1898|**deferred**| draft; missing the retry double-advance and per-account isolation tests this plan required — commented on the PR with both named |
171
+
|#1888|**blocked**|`CONFLICTING/DIRTY`, `CHANGES_REQUESTED`, and an unsponsored auth surface — three blockers, none of which an agent should clear |
172
+
173
+
Verification on the merged tree: `bun test` across
174
+
`cline-pass-deepseek-v4-tool-replay`, both `fastwire-characterization-*`, and `router` —
175
+
**54 pass, 0 fail**.
176
+
177
+
`dev` at `2a9f08324` has CI `in_progress` (run `32085152470`); the two prior dev runs were
178
+
cancelled by supersession, so the branch still has no completed green run on its current head.
179
+
That is a promotion gate for WP9, not a merge gate here.
The original text said "#1836 closes as superseded" and "#1906 stays closed." Both were
76
+
inverted and were corrected in `002_merge_order_corrections.md`; re-confirmed here at head:
77
+
**#1836 is CLOSED** already, and **#1906 is OPEN**. Nothing to do on #1836. #1906 is a genuine
78
+
open question about whether the Antigravity adapter should reach `/v1internal`, which is the
79
+
undocumented-protocol policy decision reserved for the user.
80
+
81
+
## Security posture for this wave
82
+
83
+
These PRs change how the client identifies itself upstream. Before merging either, the diff
84
+
must show no token, account id, or project value reaching a snapshot, log, or test fixture —
85
+
`AGENTS.md` treats credential handling as a release blocker, and a fingerprint change is
86
+
exactly where a capture fixture tends to acquire one by accident.
87
+
## Corrections from the WP8 audit — the order inverts, and #1891 holds
88
+
89
+
**#1891 violates this wave's own accept criterion, and I treated that criterion as a box to
90
+
tick rather than a live risk.**
91
+
92
+
*Wording corrected after review: I first called this a "leak." It is not one.* The env var is
93
+
set by whoever controls the process, and anyone who can set it can already read the token file
94
+
or patch the source. No trust boundary is crossed and no secret escapes. It is a **contract
95
+
violation and a correctness foot-gun**, and calling it a leak in a section headed "security
96
+
posture" inflates a real finding into a wrong category — which is exactly how you lose
97
+
credibility on the next finding that genuinely is severe.
98
+
99
+
The sharper objection, which I also missed: on `dev` today `ide_version` is *already* the full
100
+
UA string. The wrongness predates #1891 entirely. #1891 does not open a channel — it makes an
101
+
already-wrong channel operator-steerable.
102
+
103
+
The criterion said "a UA override never leaks into body metadata." #1891 violates it. The
104
+
change reads as consolidation — moving the `GOOGLE_ANTIGRAVITY_USER_AGENT` lookup out of the
105
+
module constant and into `antigravityUserAgent()` — but that function has an untouched caller
106
+
at `src/oauth/google-antigravity.ts:114` which puts its return value in the `onboardUser`
107
+
**request body** as `ide_version`. So the override widens from one destination to two.
108
+
109
+
Reproduced in a scratch worktree, same env var, `dev` versus `dev`+#1891:
110
+
111
+
```
112
+
baseline dev → ide_version = antigravity/ide/2.5.5 (aidev_client; os_type=windows; arch=amd64)
113
+
dev + #1891 → ide_version = LEAK-CANARY/1.0
114
+
```
115
+
116
+
**The dependency runs opposite to my reorder.** I put #1889 last because it is the only PR with
117
+
red CI. But #1889 is the PR that makes `ide_version` a real version constant — it *closes* the
118
+
hole #1891 widens. Ordering by CI colour put the fix behind the regression. The correct
119
+
sequence is: sponsor and land **#1889 first**, then #1891 becomes safe.
120
+
121
+
That does not change my refusal to self-apply `maintainer-sponsored` on #1889 — it makes the
122
+
refusal costlier, which is the honest position rather than a reason to reconsider.
123
+
124
+
### Other findings
125
+
126
+
-**#1891 adds `PI_AI_ANTIGRAVITY_USER_AGENT`**, an env var with no references anywhere else in
127
+
`src/`, `tests/`, or `docs-site/` — a second undocumented spoofing knob under a title about
128
+
token order.
129
+
-**#1891's central claim is asserted, not attached.** It cites a decompiled address and live
130
+
200s, but no disassembly excerpt or redacted capture is in the diff. For a change whose whole
131
+
value is matching an observed client, the observation is the artifact. Requested on the PR.
132
+
-**#1891 is clean on secrets** — no token, account id, or project value in the diff, fixtures,
133
+
or added tests. Checked specifically.
134
+
-**#1897 misses one of its four cache-contract requirements**: invalidation on authorization
135
+
failure. `markProviderDiscoveryFailed` neither clears the cache nor bumps the generation, so a
136
+
stale wire-ID map survives a 401/403. Incremental gap rather than regression — there was no
137
+
wire-ID cache before — so it did not hold the merge, and it is recorded on the PR.
138
+
139
+
### Corrections to this document
140
+
141
+
`#1889` has **4** failing checks at head, not 5 as the original text said. And `#1906` is an
142
+
**issue**, not a PR — the earlier correction reached the right state through the wrong object
143
+
type.
144
+
145
+
## WP8 outcome
146
+
147
+
| PR | Outcome | Evidence |
148
+
|----|---------|----------|
149
+
|#1897| merged |`aca3c0241`; **macOS-only** local verification — 99 pass / 0 fail plus `tsc` clean. No CI run existed at head, which is a fact about fork policy rather than an unavoidable constraint: pushing the head to a repo branch would have triggered `push` CI. Judged not worth it for a pure-TypeScript diff with no platform-sensitive APIs |
150
+
|#1891|**held**| makes an operator env var steerable into an upstream request body; violates this wave's accept criterion; needs #1889 first. Note its head also has **no test CI** — the four green checks are governance gates, not tests |
0 commit comments