Skip to content

v1.94.4.0 captain docs: plan-first rule matches R65 (no plan review for bug fixes) - #3908

Merged
capy-ai[bot] merged 6 commits into
mainfrom
gsta258-row8-plan-wording
Nov 3, 2026
Merged

capy-ai[bot] merged 6 commits into
mainfrom
gsta258-row8-plan-wording

Conversation

@garrytan

@garrytan garrytan commented Nov 3, 2026 •

Copy link
Copy Markdown
Owner
{
  "schema_version": 1,
  "pr": 3908,
  "head": "59bbd14c950b27c4b16498bfdede808673da4387",
  "version": "1.93.36.0",
  "base": "main",
  "gate": {
    "focused": {
      "pass": 28,
      "fail": 0,
      "skip": 0
    }
  },
  "session_kind": "unattended",
  "head_tree": "a2ec08439d337e7acb012134555221b7b2e98cff",
  "product_patch_id": "bb45a7037e698254532ac1763363e578cb75dfba",
  "gstack_version": "1.93.36.0",
  "review_depth": {
    "depth": "full",
    "class": "code",
    "runtime_lines": 31,
    "rules": "b2272efd321ce7cb",
    "base": "5192908de442e9ffd555dedfc05f993ea95b3e01",
    "registry": "5192908de442e9ffd555dedfc05f993ea95b3e01",
    "gate_scan": {
      "tagged": 0,
      "head": "59bbd14c950b27c4b16498bfdede808673da4387"
    },
    "reasons": [
      "risk:money",
      "runtime=31"
    ],
    "risk": [
      "money"
    ]
  },
  "pregate": "requires-remote cleared @a2ec08439d33 (1 checks, current for 59bbd14c950b)",
  "requires_remote": [
    "free-tests.yml/macos-named-regressions: https://github.com/garrytan/gstack/actions/runs/41568175095/attempts/1",
    "windows-smoke.yml/windows-smoke: https://github.com/garrytan/gstack/actions/runs/41568175036/attempts/1",
    "pregate.json/remote_gate: remote-gate tree=a2ec08439d33 tests=18973"
  ],
  "issues": {
    "fixes": [],
    "refs": [],
    "flakes": []
  },
  "docs": {
    "status": "current",
    "scope": "docs/, README.md, CONTRIBUTING.md, captain/, capy/skill/ (8 committed files plus docs/unattended.md and autoplan/DRIVER.md)"
  },
  "review_loop": {
    "passes": 2,
    "blocking_found": 5,
    "deferred": 0,
    "noted": 0,
    "fixed": 0,
    "scope": "frozen",
    "fix_checks": 1,
    "rounds": 3,
    "unrecorded": 5,
    "ledgers_read": 4,
    "dispatches": {
      "specialist": 2,
      "red-team": 2
    }
  },
  "review_loop_exception": {
    "accepted_by": "GStack Minder",
    "date": "2026-11-03",
    "via": "approval a-0c5c8ef7158a",
    "reason": "standing grant, GSTA-258 builds, 2026-11-03",
    "fix_check": {
      "head": "5ecc503b4206299c6b3c120b157af58a4e131190",
      "findings": 0,
      "files": [
        "captain/SKILL.md",
        "captain/SKILL.md.tmpl",
        "capy/skill/gstack/SKILL.md",
        "captain/templates/CREW.md"
      ]
    }
  },
  "review_targeted": {
    "classes": [
      "money"
    ],
    "rules": "c82addff6fb60fa0",
    "brief_sha": "7f92a0d0a8319c3c0e076fb3b1a4689c45120b2be2c14dfc7d4cb03aafa9d6a5",
    "product_patch_id": "bb45a7037e698254532ac1763363e578cb75dfba",
    "status": "done",
    "findings": 0,
    "raised_by": "targeted"
  }
}

What changed

The Capy gstack skill's plan-first rule now follows R65: bug fixes are exempt from plan review and carry a repro on current main or a failing test written first; features get the lean run. (Image render failed on this machine: PR_SHOTS_RENDER_FAILED, so this is the text version.)

Before/after as text: The Capy gstack skill's plan-first rule now follows R65: bug fixes are exempt from plan review and carry a repro on current main or a failing test written first; features get the lean run
--- Before · capy/skill/gstack/SKILL.md § Plan, approve, ship at 5192908 · source text
+++ After · capy/skill/gstack/SKILL.md § Plan, approve, ship on this branch · source text
 ## Plan, approve, ship
 Every Capy thread that uses gstack follows these for new work, under a Captain or on its own. The rule in full is `docs/captain.md`, "Plan before build"; the lane side is section 3 of captain.md in this folder.
 
+- **Plan first.** Run `/autoplan` on new functionality right after the install check. The lean run (the Capy default) is one reviewer over plan and draft code; `--full` adds every phase.
-- **Plan first.** Run `/autoplan` on new functionality right after the install check. A bug fix needs none (`gstack-autoplan next --kind fix`), just a repro or failing test in the PR. A feature gets the lean run (the Capy default), one reviewer over plan and draft code; `--full` adds every phase.
 - **ELI5, then the word.** Send the plan and gate to the person, or the Captain, in plain language: `gstack-owner-brief render --captain --out <run dir> --captain-name "<Captain name>"`, its choices as lettered lines in the ask format (`docs/captain.md`) and "do it" taking every recommendation at full scope. Build only after their word is answered into the run and it reports `APPROVED`; then `gstack-approvals import --autoplan <run dir> --by "<owner>"`.
 - **/ship before ready.** Run /ship at the final pushed head before a PR is called ready; `SHIP_RECEIPT: pr=<n> head=<head>` must name that head, and a later push needs a new /ship. The review inside /ship, like the autoplan, is sized to the change.
 - **Stop a finished /ship subagent.** When a /ship subagent opens its PR, the parent subscribes to that PR with Capy's subscribe tool and keeps that one subscription until the PR merges or closes, re-subscribing if the watch expires, then cancels it. Once the subscription is in place and READY arrives, the parent stops the subagent and owns the PR from then on, slot stamp, re-READY and receipt carry included. A stopped /ship subagent woken by a check, review or merge notice, not by its parent, reports one line naming what woke it, its pin and the PR head (`gh pr view <n> --json headRefOid --jq .headRefOid`) and ends the turn without acting, whether or not the head moved. A message from its parent is an instruction it carries out.
 - **No merge without the owner's word.** A thread never merges on its own; under a Captain the controller merges on a recorded standing approval, and a thread with no Captain asks the person.
 
+Exempt from the plan step: read-only work, the daily flaky pass's repairs, slot stamps, CHANGELOG-only resolves, cap re-pins under a written grant, S1 emergencies (a broken install or upgrade, a red main, a security bug) and fixes from an already-diagnosed, audited problem list. Other bug fixes skip plan review (`--kind fix`, R65) but carry a repro on current main or a failing test written first; features get the lean run (`--kind feature`).
-Exempt from the plan step: read-only work, the daily flaky pass's repairs, slot stamps, CHANGELOG-only resolves, cap re-pins under a written grant, S1 emergencies (a broken install or upgrade, a red main, a security bug) and fixes from an already-diagnosed, audited problem list. Every other fix gets a one-round, eng-only plan.

The same wording lands in captain/SKILL.md.tmpl (and its generated SKILL.md), captain/templates/CREW.md, captain/templates/brief.md and docs/captain.md.

Why

Build E (#3897) made gstack-autoplan next --kind fix print AUTOPLAN: not needed (bug fix; R65) and made --kind feature run one lean reviewer. Five copies of the plan-first rule still told lanes that "every other fix gets a one-round, eng-only plan", so a lane that read them would plan work the tool now skips. Now every copy says the same thing. A bug fix is exempt (the brief's reason is bug fix (R65)) and its PR carries the proof.

Blast radius

Only prose changed, in two eager-loaded skills, two crew templates and one doc. No code reads the brief's Plan: line. The eager skills stay under their context-budget ceilings: test/context-budget-ratchet.test.ts passes at 59bbd14, and it was the check that went red at 9911da7 before the wording was trimmed.

Approvals

Date Asked Exact words Outcome Covers Still pending
2026-11-03 Grant one scoped fix-check on #3908 at 5ecc503 (review_loop_exception) “standing grant, GSTA-258 builds, 2026-11-03” (GStack Minder) approved review_loop_grant: one scoped fix-check on PR #3908 at stop head 5ecc503, limited to the 5ecc503 proof-bar fix in captain/SKILL.md(.tmpl), capy/skill/gstack/SKILL.md, captain/templates/CREW.md none

Summary

Plan-first wording in the captain and Capy skills, the CREW and brief templates and docs/captain.md now matches R65. The wording was trimmed to fit the context-budget ceilings, and two review rounds tightened it. Bug fixes are named as exempt, proof means a repro on current main or a failing test written first, and the phase order is scoped to the full run.

Test Coverage

test/captain-skill.test.ts checks every copy for --kind fix and --kind feature and rejects the old "one-round, eng-only plan" phrase. Focused lane: 28 pass, 0 fail (context-budget-ratchet, captain-skill, parity-suite, captain-brief).

Pre-Landing Review

Pass 1 (full depth) had no specialist or red-team findings. The adversarial voice raised 4 notes; 3 were labeled CRITICAL by the collector and fixed in caa17fc. Pass 2 had no specialist or red-team findings; one adversarial note was fixed in 5ecc503 under the granted fix-check, which found nothing. The outside voice (openai/gpt-6.1-sol) and the targeted reader found nothing in any round.

REVIEW_LOOP: passes=2 blocking_found=5 deferred=0 noted=0 fixed=0 scope=frozen unrecorded=5 ledgers_read=4 fix_checks=1 dispatches=4

Review-loop exception: accepted by GStack Minder 2026-11-03 via approval a-0c5c8ef7158a: standing grant, GSTA-258 builds, 2026-11-03 (fix-check at 5ecc503, 0 findings)

Gate edits: none detected (GATE_SCAN tagged=0)

Review notes

  • Open note (pass 2, adversarial, Low): step 1 of the brief spells out Plan: exempt: bug fix (R65), so a lane given Plan: required could apply that exemption to itself. The controller should own that line.
  • Open note (final voices, adversarial, Low): the captain and Capy skills say "Other bug fixes skip plan review", while CREW.md and docs/captain.md say "is exempt too", and only docs/captain.md names the reason bug fix (R65).
  • Open note (pass 1, adversarial, Medium, INVESTIGATE): nothing defines where a bug fix ends and a feature begins, and --kind fix writes no record, so the label is the lane's own call.

Eval Results

evals: none affected (eval:select at origin/main: no behavioral probes, no quality judges).

Verification Results

Remote pregate at tree a2ec08439d33 (head 59bbd14): 18973 tests, 0 failed, $0.11. macos-named-regressions and windows-smoke cleared from CI runs 41568175095 and 41568175036. gstack-pregate verify: ok.

Documentation

Status: current. The R65 plan wording is consistent across every authored copy, so no doc edits were needed.

Scope: the 8 committed files in base 5192908..head 59bbd14 (none staged, unstaged or untracked), plus related docs: capy/skill/gstack/captain.md, docs/unattended.md and autoplan/DRIVER.md(.tmpl). I checked the wording against bin/gstack-autoplan and lib/autoplan-lean.ts, and searched the authored docs for the retired 'one-round, eng-only plan' and short-run fix recipe.

Documentation health:

  • docs/captain.md: Current. It matches the --kind fix / --kind feature behavior.
  • captain/SKILL.md, captain/SKILL.md.tmpl: Current, and identical to each other.
  • captain/templates/CREW.md, captain/templates/brief.md: Current.
  • capy/skill/gstack/SKILL.md, capy/skill/gstack/captain.md: Current.
  • docs/unattended.md, autoplan/DRIVER.md: Current. They already describe R65 kinds.

Coverage debt: none found. Diagram drift: none, because no diagrams are affected.

Test plan

  • bun test test/context-budget-ratchet.test.ts test/captain-skill.test.ts test/parity-suite.test.ts test/captain-brief.test.ts (28 pass)
  • Remote pregate, whole free tier (18973 tests, 0 failed)
  • CI at 59bbd14: 27 success, 0 failure, 0 pending

Issue: GST-1139 (follow-up to reports/GSTA-258-E-decisions.tsv row 8). The crew's own CREW.md copy is GStack Minder's to update.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
@trunk-io

trunk-io Bot commented Nov 3, 2026 •

Copy link
Copy Markdown

😎 This pull request was merged.

garrytan and others added 5 commits November 3, 2026 02:17
The eager-loaded captain and Capy gstack skills grew past their
context-budget ratchet ceilings (captain 6576 > 6551, capy/skill/gstack
5367 > 5315). Shorten the sentence in both, drop the duplicate bug-fix
clause from the Capy skill's Plan first bullet, and remove the stale
'short run, not no run' clause from docs/captain.md that contradicted
the new rule. The test now asks each copy for --kind fix and
--kind feature.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
… repro or failing test)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…failing test written first)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
@github-actions github-actions Bot changed the title captain docs: plan-first rule matches R65 (no plan review for bug fixes) v1.94.4.0 captain docs: plan-first rule matches R65 (no plan review for bug fixes) Nov 3, 2026
@capy-ai
capy-ai Bot merged commit 5f705f3 into main Nov 3, 2026
45 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant