@feigi/fleet-ctl 3.19.8 → 3.21.0

package/scripts/within-run-pair-prose.test.mjs

@@ -1,8 +1,7 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
-import { existsSync, readFileSync } from "node:fs";
+import { readFileSync } from "node:fs";
 import { join } from "node:path";
-import { CELL } from "./ledger-grammar.mjs";
 import { between as section } from "./prose-pin.mjs";
 
 // #864's finding is that tier is entangled with calendar date and therefore
@@ -126,14 +125,16 @@ test("phase 2 counts the rate off the ledger's impl- rows, and a replacement doe
     /records\s+`tier=<cell>`\s+in\s+the\s+row,\s+and\s+a\s+replacement\s+inherits\s+the\s+row's\s+tier/,
     "phase 2 no longer records the cell on the row, or no longer says a replacement inherits it rather than counting as a Pull",
   );
-  // Until the router draws one, the prose has to name the cell the controller
-  // writes, and that cell has to have a definition: `ledger.mjs dispatch`
-  // refuses a `tier=` naming no `agents/` file, so a cell named here without
-  // one stops every exploration Pull before its member goes live.
-  const named = /Until\s+the\s+router\s+draws\s+the\s+cell[^.]*?`<cell>`\s+is\s+`([^`]+)`/.exec(slice)?.[1];
-  assert.ok(named, "phase 2 no longer names the cell an exploration Pull writes before the router exists");
-  assert.ok(CELL.test(named), `${named} is not a cell`);
-  assert.ok(existsSync(join(REPO, "agents", `fleet-implementer-${named}.agent.md`)), `phase 2 names ${named}, which has no agents/fleet-implementer-${named}.agent.md`);
+  // The cell is the router's draw, never one this prose names: a cell
+  // written here is a second copy of the routing, free to disagree with the
+  // table the router reads.
+  assert.match(slice, /the\s+router\s+draws\s+its\s+cell/, "phase 2 no longer says the router draws the exploration cell");
+  assert.doesNotMatch(slice, /Until\s+the\s+router\s+draws/, "phase 2 names a stand-in exploration cell again");
+  assert.match(
+    slice,
+    /While\s+the\s+table's\s+`burn_in`\s+is\s+true\s+every\s+Pull\s+draws,\s+over\s+every\s+cell,\s+the\s+policy\s+cell\s+included/,
+    "phase 2 no longer says burn-in draws on every Pull, over every cell",
+  );
   // The roll: a Pull that lands on a chain head passes the exploration cell to
   // the next Pull rather than skipping it, which is the rule ADR 0013 §6 states
   // and the difficulty caveat below depends on.
@@ -144,6 +145,41 @@ test("phase 2 counts the rate off the ledger's impl- rows, and a replacement doe
   );
 });
 
+// The Pull's own write: the router's line decides the row, and `tier=` is
+// the draw's record — on a row exactly when the router drew. Written for a
+// row the router did not draw for, `ledger.mjs dispatch` would print a cell
+// nothing chose.
+const pullStep7 = () => section(RUN_TEAM, "7. **Route, row, then dispatch**", "**The Pull table**", "run-team Pull step 7");
+
+test("Pull step 7 routes through ticket-router.mjs and writes tier= exactly when it drew", () => {
+  const slice = pullStep7();
+  assert.match(slice, /ticket-router\.mjs\s+route\s+--session\s+<session>\s+--ticket\s+<N>\s+--arm\s+<A\|B>\s+--impl-row\s+<k>/, "step 7 no longer calls the router");
+  assert.match(slice, /--issue\s+<scratch>\/impl-<N>\/issue\.json/, "step 7 no longer hands the router the Pull's own read");
+  assert.match(slice, /`tier=<CELL>`\s+exactly\s+when\s+`DRAW`\s+is\s+not\s+`-`/, "step 7 no longer ties tier= to the draw");
+  assert.match(slice, /Router\s+exit\s+2\s+→\s+no\s+row\s+and\s+no\s+dispatch/, "step 7 dispatches on a router usage error");
+  assert.match(slice, /--impl-row\s+<k>\[\s+--chain-head\]/, "step 7 no longer passes --chain-head for a ticket another sequences after");
+  assert.match(slice, /--guard\s+\.fleet\/cost-guard\.json/, "step 7 no longer hands the router the cost guard");
+  assert.match(slice, /\[\s+·\s+route=<REASON>\]/, "step 7 no longer records the router's REASON on the row");
+  assert.match(slice, /release\s+the\s+claim\s+\(`release-ticket\.sh`/, "step 7 no longer releases the claim on a router exit 2");
+  // The Pull's one read is what the router reads: a second fetch would be a
+  // second network call per Pull.
+  const step3 = section(RUN_TEAM, "3. **Read the ticket in full, once**", "Record the ticket's real", "run-team Pull step 3");
+  assert.match(step3, /--json\s+title,body,comments,labels,createdAt`\s+written\s+to\s+`<scratch>\/impl-<N>\/issue\.json`/, "step 3 no longer writes the file the router reads");
+});
+
+const routerArgs = () =>
+  section(RUN_TEAM, "**The router's arguments, and nothing past its line.**", "**Count the rate against the ledger", "run-team router's arguments");
+
+test("the router's arguments paragraph pins the roll, the arm parity and the single source of CELL", () => {
+  const slice = routerArgs();
+  assert.match(slice, /row\s+5k\s+and\s+every\s+row\s+after\s+it\s+carry\s+no\s+`tier=`/, "the --impl-row rolling definition is gone");
+  assert.match(slice, /`--chain-head`\s+marks\s+a\s+ticket\s+another\s+open\s+ticket\s+sequences\s+after/, "--chain-head is no longer defined");
+  assert.match(slice, /`A`\s+for\s+an\s+even\s+ticket\s+number\s+and\s+`B`\s+for\s+an\s+odd/, "the --arm parity rule changed");
+  assert.match(slice, /never\s+re-derive\s+`CELL`/, "the router's line is no longer the whole decision");
+  assert.match(slice, /`REASON=guard-missing`/, "the guard-missing route is no longer named");
+  assert.match(slice, /`REASON`\s+other\s+than\s+`ok`\s+goes\s+on\s+the\s+row\s+as\s+`route=<REASON>`/, "REASON no longer lands on the row as route=");
+});
+
 // The difficulty caveat lives in the counter-evidence section, past this file's
 // dispatch slice, so it needs its own anchor pair.
 const counterEvidence = () =>

package/skills/run-team/SKILL.md

@@ -515,9 +515,10 @@ terminating in exactly one dispatch or an empty shortlist (ADR 0013 Decision 2):
    nothing else; work done directly against the shared checkout, by a
    human or another agent session, stakes no branch-shaped claim and trips
    none of them.
-3. **Read the ticket in full, once** — `gh issue view <N> --json title,body,comments,labels
-   --jq '.title, .body, ([.labels[].name]|join(",")), (.comments[]|.author.login + ": " + .body)'`.
-   One read answers every question the Pull asks. Record the ticket's real
+3. **Read the ticket in full, once** — `gh issue view <N> --json title,body,comments,labels,createdAt`
+   written to `<scratch>/impl-<N>/issue.json`, then read off that file with `jq -r
+   '.title, .body, ([.labels[].name]|join(",")), (.comments[]|.author.login + ": " + .body)'`.
+   One read answers every question the Pull asks; step 7's router reads the same file. Record the ticket's real
    brief — the `## Agent Brief` comment where one exists, otherwise the issue
    body, which is where most tickets actually carry it (a controller probe found only 43/100 open
    `ready-for-agent` issues have a separate Agent Brief comment; the rest,
@@ -706,9 +707,16 @@ terminating in exactly one dispatch or an empty shortlist (ADR 0013 Decision 2):
    two independent tickets.
 6. **Claim** — `~/.fleet/bin/fleet-run claim-ticket.sh <N> impl-<N> implementer
    --apply` (below). Exit 2 → treat it as taken, next entry.
-7. **Row, then dispatch** — `ledger.mjs row <N> "impl-<N>[ · tier=<cell>]"`,
-   then phase 2 at once: claim what you are about to dispatch;
-   dispatch what you have just claimed.
+7. **Route, row, then dispatch** — `~/.fleet/bin/fleet-run ticket-router.mjs route
+   --session <session> --ticket <N> --arm <A|B> --impl-row <k>[ --chain-head] --issue
+   <scratch>/impl-<N>/issue.json --guard .fleet/cost-guard.json` prints
+   `POLICY=… CELL=… DRAW=… STRATUM=… REASON=…` (phase 2's exploration rule
+   says what each argument is); then `ledger.mjs row <N> "impl-<N>[ · tier=<CELL>][ · route=<REASON>]"`,
+   `tier=<CELL>` exactly when `DRAW` is not `-`; then phase 2 at once: claim
+   what you are about to dispatch; dispatch what you have just claimed. Router
+   exit 2 → no row and no dispatch: release the claim (`release-ticket.sh`,
+   below) and emit an event, a probe that could not look — a table the router
+   refuses refuses every Pull.
 
 **The Pull table** — every verdict a Pull can reach, and the one thing each
 does (ADR 0013 Decision 2 and 4):
@@ -923,8 +931,8 @@ kernel and needs a `wait`, while a `task` job is process-level and its report
 auto-delivers on its own, like every other member's.
 
 **No per-call tier or effort anywhere on this path**: tier stays the agent
-definition's own frontmatter (ADR 0005), whichever of the two definitions the
-Pull names, and a member's Resolved tier stays readable from its transcript.
+definition's own frontmatter (ADR 0005), whichever `fleet-implementer-<cell>`
+definition the Pull names, and a member's Resolved tier stays readable from its transcript.
 
 **Claim what you are about to dispatch; dispatch what you have just claimed.**
 Phase 1 makes the claim, serially in the main checkout — its rule is measured —
@@ -1087,16 +1095,31 @@ dispatched — the fleet reads that config and never writes it (ADR 0003).
 
 **Every 5th Pull by ledger count goes at an exploration cell.** Count the `impl-`
 rows in `.fleet/ledger.md` at Pull time; the Pull that creates row 5, 10, 15 …
-records `tier=<cell>` in the row, and a replacement inherits the row's tier —
+is an Exploration Pull, and the router draws its cell: `k = 1 +
+(sha256("<session>\t<ticket>")[0:8] mod K)` over `router-table.json`'s
+`cells` minus the policy cell, in token order. While the table's `burn_in` is
+true every Pull draws, over every cell, the policy cell included. An
+Exploration Pull records `tier=<cell>` in the row, and a replacement inherits the row's tier —
 written at the Pull's step 7, ahead of `ledger.mjs dispatch`, which reads it
-and prints `fleet-implementer-<cell>` as the `agent` to dispatch. Until the
-router draws the cell (spec 2026-09-28 § 2), `<cell>` is `task-high` — the
-retired alternate tier's own `@task:high` route. The
+and prints `fleet-implementer-<cell>` as the `agent` to dispatch; a row with
+no `tier=` runs the policy cell. The
 assignment rolls to the next Pull when another open ticket sequences after
 it — a ticket the rest of the run
-depends on. Do not tell the member it is a control: a member that
+depends on; `class` is not read. Do not tell the member it is a control: a member that
 knows it is being measured is not measuring the same thing.
 
+**The router's arguments, and nothing past its line.** `--impl-row` is the
+row count the Pull creates — or, while an assignment is rolling (the latest
+row 5k and every row after it carry no `tier=`), that row's number;
+`--chain-head` marks a ticket another open ticket sequences after, which is
+what rolls it. `--arm` is `A` for an even ticket number and `B` for an odd
+one; a B Pull routes exactly as an A one until the table opens arm B.
+`--session` is this run's own session directory. The printed line is the
+whole routing decision — never re-derive `CELL` from the table, the guard or
+this paragraph. A missing or unreadable `.fleet/cost-guard.json` routes the
+default cell only (`REASON=guard-missing`), burn-in included, and `REASON`
+other than `ok` goes on the row as `route=<REASON>`, nowhere else.
+
 **Count the rate against the ledger, because a Pull is the dispatch unit.** One
 Pull is one member, so a rate counted per Pull is a rate counted per member —
 one implementer in five, however the Pulls fall across the run — and the
@@ -1353,31 +1376,35 @@ awk -F'\t' '!/^#/ && $4=="routine" {t[$5]++; if($6=="no") f[$5]++} \
 ```
 
 **These rows are the historical corpus, and nothing appended to them fixes the
-confound.** What fixes it is the within-run pairing in the dispatch rule above:
-from now on one implementer in every five runs at the alternate tier beside the
-top-tier ones, against the same prompts on the same day, so the comparison stops depending on
-which tier history happened to leave in which week. Read the pairs, not the
-whole-file split, once there are enough of them.
+confound.** What fixes it is the exploration draw in the dispatch rule above:
+one implementer in every five — every one, during burn-in — runs at a drawn
+cell beside the policy cell's, against the same prompts on the same day, so
+the comparison stops depending on which tier history happened to leave in
+which week. Read each cell's comparisons, not the whole-file split, once there
+are enough of them.
 
 **So read what the guard actually established: routine tickets sometimes fail to
 close their own ticket. Not that `sonnet` caused it.** The revert is the
 pre-committed rule being honoured, not a measurement. Restoring a cheap tier is
 still the maintainer's call and still a change to the dispatch rule — but the
 deliberate control it used to require is no longer something anyone has to
-authorize one ticket at a time: the alternate-tier dispatch above produces one
-in every five Pulls by construction. **Do not read the pairs early.** Report the count —
-the DELIBERATE count, from the query in `member-outcomes.tsv`'s header, which
-prints the pair count and its distinct `run_date`s and counts only sessions
-that ran both implementer definitions at different models — and stop until
+authorize one ticket at a time: the exploration draw above produces one in
+every five Pulls by construction. **Do not read a cell early.** A cell's
+comparison is one session holding an admissible row at that cell — its tier
+check passed and its resolved effort is the cell's level — and one at
+`slow-high` whose resolved model and effort differ. Count them per cell, never
+pooled across cells, and stop until
 there are at least ten of them across five or more distinct `run_date`s; below
-that, a pair count is a number, not evidence, and the last guard fired with
-n=1 on the control side. **Report whichever number that query prints, never a
-number from any other one:** the query it replaces (#1066) counted every
-session whose implementers merely differed, so it cleared this floor by an
-order of magnitude while the controlled comparison did not exist yet.
-
-Why the agent body carries what it does — read this before editing either agent
-file, and keep the two bodies byte-identical. Each rule in the body's
+that, a comparison count is a number, not evidence, and the last guard fired
+with n=1 on the control side. **Report the count the per-cell readout prints,
+never one from any other query:** until that readout exists there is no count
+to report, the pair query in `member-outcomes.tsv`'s header counts pre-cutover
+pairs only, which are no cell's comparisons, and the query before that one
+counted every session whose implementers merely differed, so it cleared this
+floor by an order of magnitude while the controlled comparison did not exist yet.
+
+Why the agent body carries what it does — read this before editing any
+`fleet-implementer-<cell>` file, and keep every cell's body byte-identical. Each rule in the body's
 enumerate-and-declare block is load-bearing, for a different reason.
 **Enumerate-and-declare** answers a signature measured four times in one run:
 each implementer fixed exactly the cases its ticket named and left an adjacent
@@ -3667,7 +3694,7 @@ naming the token and writing nothing, because every reader would take it as a
 permanent settle.
 
 `→ PR#M` stays as the human-readable arrow; the tick reads only the `=`
-tokens. `tier=<cell>` marks the every-5th-Pull member's cell (phase 2), and `excluded ·
+tokens. `tier=<cell>` marks an Exploration Pull's cell and `route=<REASON>` a router fallback (phase 2), and `excluded ·
 behind-pr:#M | behind-issue:#M` is an Exclusion (phase 1) — a ticket row like
 any other, never a section of its own.