Say When an Earlier Partial Round Exits 42 in pr_review Docs

.agents/skills/pr-review-conduct/SKILL.md

@@ -45,11 +45,13 @@ visible comments, routinely still carries a finding nobody has answered. Treatin
    by commit SHA rather than assumed from a green merge-state. A push makes checks go green *before* the
    re-review lands, and the matched review is **read**, not just counted. A review can carry the
    head SHA and still decline the PR outright, or say it read only part of the changed files.
-   `pr_review.py` reads coverage into the digest's `coverage=` field, and only a stated partial
-   is an exit code, `42`. A round covering the head that states full coverage satisfies this
-   item. Where none states any, a round covering the head whose own file table names exactly the
-   changed files covers it, `coverage=table`, the reading a Copilot round at Balanced review
-   effort gives. Where no round covering the head carries a table of its own, the newest round
+   `pr_review.py` reads coverage into the digest's `coverage=` field, and only partial coverage
+   is an exit code, `42`. That is a head round stating partial, or a head round reading `UNKNOWN`
+   beside an earlier partial round. So is a head with non-refusal rounds, none stating coverage,
+   beside one. A round covering the head that states full coverage satisfies this item where no
+   other head round reads `UNKNOWN` or partial. Where none states any, a round covering the head
+   whose own file table names exactly the changed files covers it, `coverage=table`. That is the
+   reading a Copilot round at Balanced review effort gives. Where no round covering the head carries a table of its own, the newest round
    that does, its table naming exactly the changed files, stands in as `coverage=carried:table`.
    A table stands in only where no Copilot round on the pull request, on any commit, states
    partial coverage, so a pull request that ever had a partial round goes to the maintainer.

.claude-plugin/fleet-skills/.source-digests/pr-review-conduct

@@ -1 +1 @@
-75b92a3bbf15ab34
+184af41cfc570242

.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md

@@ -45,11 +45,13 @@ visible comments, routinely still carries a finding nobody has answered. Treatin
    by commit SHA rather than assumed from a green merge-state. A push makes checks go green *before* the
    re-review lands, and the matched review is **read**, not just counted. A review can carry the
    head SHA and still decline the PR outright, or say it read only part of the changed files.
-   `pr_review.py` reads coverage into the digest's `coverage=` field, and only a stated partial
-   is an exit code, `42`. A round covering the head that states full coverage satisfies this
-   item. Where none states any, a round covering the head whose own file table names exactly the
-   changed files covers it, `coverage=table`, the reading a Copilot round at Balanced review
-   effort gives. Where no round covering the head carries a table of its own, the newest round
+   `pr_review.py` reads coverage into the digest's `coverage=` field, and only partial coverage
+   is an exit code, `42`. That is a head round stating partial, or a head round reading `UNKNOWN`
+   beside an earlier partial round. So is a head with non-refusal rounds, none stating coverage,
+   beside one. A round covering the head that states full coverage satisfies this item where no
+   other head round reads `UNKNOWN` or partial. Where none states any, a round covering the head
+   whose own file table names exactly the changed files covers it, `coverage=table`. That is the
+   reading a Copilot round at Balanced review effort gives. Where no round covering the head carries a table of its own, the newest round
    that does, its table naming exactly the changed files, stands in as `coverage=carried:table`.
    A table stands in only where no Copilot round on the pull request, on any commit, states
    partial coverage, so a pull request that ever had a partial round goes to the maintainer.

.github/skills/pr-review-conduct/SKILL.md

@@ -45,11 +45,13 @@ visible comments, routinely still carries a finding nobody has answered. Treatin
    by commit SHA rather than assumed from a green merge-state. A push makes checks go green *before* the
    re-review lands, and the matched review is **read**, not just counted. A review can carry the
    head SHA and still decline the PR outright, or say it read only part of the changed files.
-   `pr_review.py` reads coverage into the digest's `coverage=` field, and only a stated partial
-   is an exit code, `42`. A round covering the head that states full coverage satisfies this
-   item. Where none states any, a round covering the head whose own file table names exactly the
-   changed files covers it, `coverage=table`, the reading a Copilot round at Balanced review
-   effort gives. Where no round covering the head carries a table of its own, the newest round
+   `pr_review.py` reads coverage into the digest's `coverage=` field, and only partial coverage
+   is an exit code, `42`. That is a head round stating partial, or a head round reading `UNKNOWN`
+   beside an earlier partial round. So is a head with non-refusal rounds, none stating coverage,
+   beside one. A round covering the head that states full coverage satisfies this item where no
+   other head round reads `UNKNOWN` or partial. Where none states any, a round covering the head
+   whose own file table names exactly the changed files covers it, `coverage=table`. That is the
+   reading a Copilot round at Balanced review effort gives. Where no round covering the head carries a table of its own, the newest round
    that does, its table naming exactly the changed files, stands in as `coverage=carried:table`.
    A table stands in only where no Copilot round on the pull request, on any commit, states
    partial coverage, so a pull request that ever had a partial round goes to the maintainer.

scripts/README.md

@@ -203,7 +203,7 @@ uv run --no-project --script scripts/pr_review.py cleanup 452 --repo ptr727/Proj
 | `30` | `wait` | Still pending at the timeout, which is pending rather than failed. On a held head, `status=CHECKS_PENDING`. A round past `--min-rounds` not yet landed. |
 | `40` | `wait` | No Copilot review covers the head: the reviewer answered outside a formal review, the request returned success and recorded nothing, or the reviewer writes under a login no query filters on. |
 | `41` | `wait` | The newest round is a refusal or a quota state, its kind printed in the status line, and in `refusal=` for a refusal on this pull request. |
-| `42` | `status`, `wait` | The round covering the head states partial coverage, or states none while an earlier round states partial coverage. |
+| `42` | `status`, `wait` | The round covering the head states partial coverage. Or a non-refusal head round exists while an earlier round states partial, and one reads `UNKNOWN` or none states coverage. |
 | `44` | `wait` | The review loop closed, the merge reads `BLOCKED`, and a required check is in a shape no wait clears. |
 | `49` | `wait` | No Copilot round covers the head and none was requested, on a pull request into a branch other than the default with no attestation. |
 | `50` | `status`, `wait` | The round covering the head opens on a verdict other than the clean one while it counts no finding. Or a head round carries a part the reader cannot place, and nothing acknowledges its fingerprint. |
@@ -224,7 +224,7 @@ uv run --no-project --script scripts/pr_review.py cleanup 452 --repo ptr727/Proj
 
 The digest reports the completed head review's effective effort as `lite`, `balanced`, or `max`, with `effort_source=default` for `Default (<level>)` and `explicit` for a bare level, and `unknown` for both where the body states none. Effort never changes a verdict, and the workflow never selects or changes the user-controlled setting.
 
-**`coverage=` names how the head's diff is shown read, and only a stated partial is an exit.** `full` and `PARTIAL` are the reviewer's own statement, the `fleet-review` marker the fleet's review instructions ask for, and `UNKNOWN` is a statement this cannot believe, such as one reading more files than changed, or a body the reader failed on, the reason printed beside it. A count Copilot states in a top-level paragraph that opens on it, never a list item and never inside a counted finding section, is read too, and a partial wins over any other reading beside it. A head round stating nothing while an earlier round states partial reads as `PARTIAL`, so a pull request that ever had a partial round goes to the maintainer. Where nothing states coverage, a round covering the head whose own file table names exactly the changed files reads as `table`, and the newest round on any commit whose table does reads as `carried:table`. A table stands in only where no round states partial coverage and the review history is not cut by its window, since a table names the whole changed set on a partial round too. The changed-file list is paged to GitHub's own ceiling of 3,000 files, and a list not read whole, or read empty, is read again before a table is matched against nothing. A table row matches a changed path exactly, or by a `*` or `?` glob. Otherwise `coverage=none` prints with the reason no table stood in, `status` adds a note, and the exit stays `0`, since Merge Gate item 2 in `pr-review-conduct` reads this field rather than an exit code. A partial round is reported and handed over rather than retried into, since every partial on record stayed partial at the identical ratio across every later round.
+**`coverage=` names how the head's diff is shown read, and only partial coverage is an exit.** `full` and `PARTIAL` are the reviewer's own statement, the `fleet-review` marker the fleet's review instructions ask for, and `UNKNOWN` is a statement this cannot believe, such as one reading more files than changed, or a body the reader failed on, the reason printed beside it. A count Copilot states in a top-level paragraph that opens on it, never a list item and never inside a counted finding section, is read too, and a partial wins over any other reading beside it. Beside an earlier partial round, a head round reading `UNKNOWN` reads as `PARTIAL`, and so does a head with non-refusal rounds, none stating coverage. Where nothing states coverage, a round covering the head whose own file table names exactly the changed files reads as `table`, and the newest round on any commit whose table does reads as `carried:table`. A table stands in only where no round states partial coverage and the review history is not cut by its window, since a table names the whole changed set on a partial round too. The changed-file list is paged to GitHub's own ceiling of 3,000 files, and a list not read whole, or read empty, is read again before a table is matched against nothing. A table row matches a changed path exactly, or by a `*` or `?` glob. Otherwise `coverage=none` prints with the reason no table stood in, `status` adds a note, and the exit stays `0`, since Merge Gate item 2 in `pr-review-conduct` reads this field rather than an exit code. A partial round is reported and handed over rather than retried into, since every partial on record stayed partial at the identical ratio across every later round.
 
 **Findings no thread carries are counted and printed whole.** `suppressed=`, `previously_missed=`, and `cr_outside_diff=` count the findings a review body collapses into a counted section, read over every round rather than the head's alone, since a push does not answer a finding nobody replied to. Each reads `N (on_head=X earlier=Y)` with a trailing `+` where the reviews window was cut. A tally holding a count the reader cannot believe reads `?`. `previously_missed=` and `cr_outside_diff=` print only where one was raised or the window was cut. `qodo_open=` counts the entries in Qodo's findings comment carrying neither its `Resolved` nor `Dismissed` badge, `unknown` where that comment can sit behind the window. `overview=T/M` is the open findings the round covering the head states beside the threads it opened, `?` where it states no count. A count it cannot believe reads `?` too. `FINDINGS WITH NO THREAD` names findings the round names, in its count or its findings list, that no open entry links to a thread on the pull request. None of these decides an exit code, and each takes the triage a suppressed finding gets, answered with `comment` since no thread exists to reply on.