Repository navigation
Folder: a request that the person does something goes through the question tool too (root 0211 addendum, base-instructions-74) - #619
Conversation
agentrozjedemeai
left a comment
There was a problem hiding this comment.
@immakermatty — Exact-head review of 3e77483. The new question rule matches root 0211 addendum in HumanAndMachines/Lazurio#623, uses Czech singular address, and preserves the rest of the rule. Buddy's generated .hermes.md is 16,665 characters (cs) / 18,209 (en), below 20,000. The two edited snapshot files change only revision/rule lines.
Blocking: the generated AGENTS.md change also invalidates eight golden snapshots in tests/snapshots/try-overlay-preset.test.ts.snap, but this file was not updated. CI tests (2/2) fails, and I reproduced it locally with bun test --max-concurrency 1 tests/try-overlay-preset.test.ts: 14 pass / 8 fail; tests/try-overlay-preset.test.ts:99 and :121 compare stale AGENTS.md SHA-256 values. The rest of each mismatch is unchanged. Please update only the eight expected AGENTS.md digests after verifying their generated text, rerun the targeted test and full CI on a new head, then request a fresh exact-head review. Folder suite (372 pass), lint, typecheck, check:public and diff --check pass on this head; they do not cover this failing golden test. No merge or deployment performed.
…n rule (Pablo's review of #619) The eight golden snapshots of tests/try-overlay-preset.test.ts pin the digest of each golden Environment's AGENTS.md: henry-01, lazurio-bob's Work Environment, services-01, team-01 and the four try-01 recordings. The new question rule changes that text, so those eight digests change, and only those. I checked the generated text against the folder snapshots: the one changed line is the rule. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
agentrozjedemeai
left a comment
There was a problem hiding this comment.
@immakermatty Approved on exact head abb1bbe834505583d5b91264c324c58c63c4c4f6. The follow-up changes only the 16 Czech/English AGENTS.md digest entries across the eight golden Environment snapshots; no other snapshot key changed. The generated question-rule text is covered in both locales. In a fresh detached checkout I ran bun test tests/try-overlay-preset.test.ts (22 pass, 8 snapshots), bun test tests/folder (372 pass, 42 snapshots) and git diff --check (clean). Current-head CI static checks and both test shards are green; zero unresolved threads. No merge, push or release.
|
Heads-up on the revision collision: #621 merged at 18:09Z (squash baa6eda) and claimed
#621 also changed the generated texts around Publication (no Lazurio "Draft"), the access identity line and the new Git and GitHub section of |
…n rule (Pablo's review of #619) The eight golden snapshots of tests/try-overlay-preset.test.ts pin the digest of each golden Environment's AGENTS.md: henry-01, lazurio-bob's Work Environment, services-01, team-01 and the four try-01 recordings. The new question rule changes that text, so those eight digests change, and only those. I checked the generated text against the folder snapshots: the one changed line is the rule. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
abb1bbe to
ebefbff
Compare
agentrozjedemeai
left a comment
There was a problem hiding this comment.
Approved at exact head ebefbff2fd70133126d82718c3c75b6906d3b1f6 against main baa6eda175a8bb5f30aa3dfabee8d2ef54031dd0. Re-reviewed both retained commits after the rebase: revision 73 and its rendered digest pin, the two tools assertions, and both ordered F14 addenda (#621 is preserved). Compared the generated snapshots against main: manual changes are revision-only; render changes are revision plus the new Czech/English question sentence; the overlay changes only the 16 AGENTS.md golden digests. Local Bun 1.4.4 validation: lint, typecheck, 177 targeted tests (107 + 70), all passing; CI static and test jobs are green. No unresolved review threads. This review does not cover the separately stacked PR #623.
…stion tool too (root 0211 addendum, base-instructions-74) Root decision 0211, addendum of 2026-10-24 (HumanAndMachines/Lazurio#623): when an agent needs the person to do something, such as entering a verification code, signing in, confirming or checking something, it asks with the harness's question tool. Matěj: the question then pops up as a notification in T3 Code; a request written only into a message gets missed, and that day a sign-in code expired. `questionTool`, which every Folder's AGENTS.md and Buddy's `.hermes.md` carry, now says so. The question gives the exact step with the link, the code and until when it is valid, and its options say how it went. The rest of the rule is unchanged. Template revision base-instructions-74 with its digest. The snapshots change only in the revision line and the rule line. The F14 addendum records it, and folder-manual.test.ts checks the new sentences in both locales. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…n rule (Pablo's review of #619) The eight golden snapshots of tests/try-overlay-preset.test.ts pin the digest of each golden Environment's AGENTS.md: henry-01, lazurio-bob's Work Environment, services-01, team-01 and the four try-01 recordings. The new question rule changes that text, so those eight digests change, and only those. I checked the generated text against the folder snapshots: the one changed line is the rule. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ebefbff to
df10b21
Compare
agentrozjedemeai
left a comment
There was a problem hiding this comment.
@immakermatty — Exact-head review of df10b21 against b5c6521: approved for content. Both commits are retained. The only functional change from the previously reviewed head is the template revision 73 → 74 and its pinned digest/assertions; the F14 addendum from #626 remains untouched. I compared the three golden snapshot files to main after normalizing only the revision line, the new cs/en question sentence and AGENTS.md digests: no residual differences, including no changes to #626 output. In an isolated checkout, lint, typecheck, check:public and all 206 focused tests passed (52 snapshots); git diff --check is clean.
CI is NOT green on this head: tests (2/2) failed on two unrelated public-vault browser tests (Chrome did not start / timeout), so the ubuntu aggregate check failed. This approval is not a merge-ready claim. Please rerun/resolve the failing CI job; do not merge until all required checks are green on this exact head. No merge, push or release performed.
…nal v0.1.11 (DEV-6656, DEV-6671) (#638) * Qualification: the evidence of v0.1.11-rc.6's canary and why the final tag stands on a release branch (in progress) The canary of v0.1.11-rc.6 ran 2026-10-24T21:15:38Z to 2026-10-25 ~05:2xZ on the pilot Organization's four work VMs and one operator's personal VM. This file records it and explains why the final v0.1.11 is tagged on release/v0.1.11 (the candidate's tree on top of the commit that merges the record) and not on main, which received #619 and #623 after the candidate. The canary record follows in this pull request once the personal VM's end readback is in. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Qualification: the canary record of v0.1.11-rc.6, 2026-10-24T21:15:38Z to 2026-10-25T05:22:09Z 8 h 6 min 31 s on the pilot Organization's four work VMs and one operator's personal VM. Every Machine was on 0.1.11-rc.6 with `lazurio recover` healthy at the start and the end and its Launchpad unit at NRestarts 0. `scripts/qualify/check-canary.ts` (offline): "The canary of v0.1.11-rc.6 passed: 8 hours on every listed Machine." The evidence file gets its end times and per-Machine table. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Qualification: the G.3 observations inside the stage, and the work VMs' window extended to 05:42:02Z Pablo's review of #638 (112c94e) found that the evidence lacked in-stage G.3 observations. Instead of inferring them, the work VMs' stage was extended and the missing work was done inside it: - a Folder refresh plus one deliberate Launchpad restart on one work VM (05:40:19Z, back healthy on the candidate, NRestarts 0); - module starts under the candidate on the three work VMs that hold modules (a built App and a development run); - per-Machine NRestarts and the Launchpad's health version at the end, with the units' active-since timestamps showing no restart of their own during the stage; - the issue search. The record's end moves to 2026-10-25T05:42:02Z (8 h 26 min 24 s). The personal VM's own window ends at 05:22:09Z (8 h 6 min 31 s), before a planned change of its lane; the evidence says so and reads nothing of it after that. The evidence no longer says the final tag carries "exactly" the candidate's tree: it carries that tree plus the two evidence files the release cycle allows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Qualification: the personal VM's window is established by its Launchpad unit's journal Pablo's review of #638 (81434c4) accepted the personal VM's separate window but asked for evidence of its Launchpad's uninterrupted run over that window, not an end counter alone. The Environment's user read the unit's journal on that VM at 2026-10-25T10:28:53Z. The only start/stop events between 20:20Z and 05:23Z are at 20:31:35Z, during the candidate's apply and before the stage; nothing falls inside 21:15:38Z–05:22:09Z. No module app unit started or stopped in the window. The evidence records this, makes the health relationship explicit (`recover` healthy requires `/health` to equal the active version, `judgeHealth`), and gives the personal VM's module scope as not applicable, with the reason. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Qualification: the evidence names every pull request that reached main after the candidate #633 (f94594e) merged into main at 2026-10-25T10:31Z, after the evidence was written, so the count "two pull requests" was no longer true. The evidence lists the pull requests by the time the record was approved and says that they and anything later go into v0.1.12. Text only; the record is unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Status: ready for review
Why
Root decision 0211, addendum of 2026-10-24 (HumanAndMachines/Lazurio#623). Matěj asked that when an agent needs the person to do something, such as enter a verification code, it asks with the question tool, so the request pops up as a notification in T3 Code. A request written only into a message gets missed: that day a sign-in code expired.
What changes
questionTool(src/folder/render.ts) gains one passage in both locales: ask the same way when the user has to do something (enter a verification code, sign in, confirm or check something). The question says exactly what to do, with the link, the code and until when it is valid, and its options say how it went ("Hotovo", "Nejde to"). A question pops up as a notification in T3 Code and in Codex Desktop; text in a message does not.AGENTS.mdand Buddy's.hermes.md, both of which usequestionTool.base-instructions-74(Folder: the agent's work is a pull request, GitHub is the authority for source code, and every agent gets Git, fork and safety rules (base-instructions-72, DEV-6617) #621 took 72 and Modules: the Bun an App pins is its runtime, resolved for every path that runs its code (F62, #618) #626 took 73):tests/folder-tools.test.tsare updated;tests/try-overlay-preset.test.tschange only in theirAGENTS.mddigest (Pablo's review).docs/decisions.md.tests/folder-manual.test.tschecks the new sentences in both locales.What does not change
Rollout
The change reaches Environments with the next Platform release and their normal Folder refresh, together with #613 and #614 (base-instructions-70 and 71).
Verification
bun test tests/folder: 372 pass, 0 fail.bun run lint,bun run typecheck,bun run check:publicandgit diff --check: clean.Plan: DEV-6676. Claude Code thread
8508a2b1-a641-4479-a132-0b188866ce6a.Rebase of 2026-10-24 ~18:45Z: rebased onto main
baa6eda1(#621). Both commits are kept. The revision is nowbase-instructions-73with its new digest; the decision record keeps #621's addendum and adds this one after it. The Folder and try snapshots differ from main only in the revision and the question rule's sentence, and in the golden Environments'AGENTS.mddigests. Local checks: lint, typecheck and 177 tests in nine Folder/try/public files pass. Pablo's approval was forabb1bbe8, so the new head needs a new verdict.Second rebase, 2026-10-24 ~22:25Z: onto main
b5c6521c(#626, which tookbase-instructions-73). The revision is nowbase-instructions-74with its new digest. Both commits are kept: the rule with the Folder snapshots, then the golden Environments'AGENTS.mddigests. Against main, the snapshots differ only in the revision line, the question rule's sentence and those digests. Local checks: lint, typecheck, check:public and 206 tests in ten Folder/try/public files pass. Pablo approvedebefbff2, so the new headdf10b216needs a new verdict.🤖 Generated with Claude Code