Repository navigation
Integration batch P (9 PRs) - #1464
Merged
Merged
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Undo Close passed only the explicit tldrIdentity, so an agent whose identity was derived (reporting domains on, no explicit field) came back with a fresh one. It now uses tldrIdentityForSession, the rule Reload Agents uses. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…bers Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…d-identity # Conflicts: # src/renderer/src/workspace/hook/actions/undoClose.ts
…d-identity # Conflicts: # src/renderer/src/workspace/hook/actions/undoClose.ts
…ion child (#1370) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…fter the deadline (#1370) A healthy provider can take longer than the bridge's 30 s create deadline to start (the bundled OpenCode took 43.1 s under load). The caller got "outcome unknown - do not repeat", the tool handler returned without delivering the prompt, and the late answer was adopted as an idle child with no brief and nothing telling the parent. - The tool handler hands the bridge a late-create continuation; the bridge keeps it for the outstanding request and runs it from adoptLateResponse (never for an in-time or failed answer). A throw is recorded as a prompt_delivery_failed incident. - The handler's bootstrap delivery is one function shared by the punctual and late paths (deliver, armPromptWhenReady, mark delivered). - The outcome-unknown reply says a late child receives its prompt automatically and must not be sent it again. Fail-first: bridge and real-MCP-handler tests are red on origin/main. Mutations killed: adoption skipping the continuation, the handler not registering it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ch, duplicate and rewind (#927) All three operations ran a real native-resume projector and threw its report away, so a projection that dropped or demoted content was invisible whenever the context reduction reported no shrink. Each result now carries a bounded, curated NativeProjectionFidelity summary (counts, per-code counts, capped source lines with an exact omitted count; no projector prose or evidence), through IPC to the renderer. The switch toast, the bulk summary and the duplicate command name material loss (UNCONFIRMED default: non-opaque drops and demotions). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…rojection-fidelity
… first in toasts, narrow the failure claim (#1384 round 1, steering q90) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…rojection-fidelity
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…prompts' (#1306) Every source (Claude, Codex, OpenCode, Pi, Grok) raises a typed ConversationPromptsUnreadable when a transcript or store is there but cannot be read or listed; only genuine absence answers []. promptFolder rethrows non-ENOENT errors; View Prompts says it in fixed words; search degrades that one row. Steering q116: discovery (stat/readdir walks) follows the same rule. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t, and Update now says a rejected request (#1425) runUpdate caught nothing after publishing the undismissable 'updating': openLog's mkdir (EACCES/ENOSPC) stranded it forever. It now publishes failed { reason: 'could-not-start', logPath: null } without running the command. The banner says so in fixed words with Update now as the retry (keyed per attempt), open-log opens nothing for it, and every Update now action awaits its IPC and shows 'Couldn't start the update. Try again.' on rejection instead of dropping it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…MCP host stop (#1372) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…out a Resume that fails (#1348) A null snapshot (run-not-found) became a generic error with Retry, and the view fell back to the reference's launch-time status, so a failed reference kept a Resume that could only fail. A new 'missing' store phase and one shared wording: Expired when the reference names its own project, Unavailable when it was looked up in the session's. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…the OS text (#1427) safeErrorMessage passed Node's own text through, so the Skills, Conventions and Custom Skills surfaces showed 'EACCES: permission denied, mkdir <home>/.claude/skills' and the like. One mapper replaces its four copies and three direct error.message sites: a system error becomes a fixed sentence per code, the service's own curated text is kept unless it carries an absolute path or runs long, anything else is generic, and the raw error goes to the main log. Real-filesystem tests (read-only skills folder, state file under a file) are red on main; the pass-through mutant is red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…wner-approved (B6 proxy) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…s package's own single-line text (#1456 review a) Review a: (1) the four Reveal handlers returned shell.openPath's raw string with the path; (2) unclassified git and fetch errors passed through because the mapper kept any short message; (3) the path filter missed /tmp, ~/, ../, quoted paths and UNC shares. Reveal failures now answer in fixed words (raw text logged); git, fetch, YAML parse and toJS failures get fixed sentences at the source; the mapper keeps only single-line, bounded, path-free text from exactly a plain Error or GitHubSkillSourceError, never child-process or library errors, and every rewrite is logged. The renderer IPC-rejection half of finding 4 is filed as #1459. All four Reveal handlers red on the old code; all five mutants red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… every errno sentence and every rewrite's log (#1456 review c) Review c: 'Provider failed [/Users/...]' was kept because '[' was not a listed delimiter; PATH_LIKE now treats any non-word character before the slash as a path start, and prose like and/or stays kept. All 18 errno codes are pinned to their sentences and each asserts its raw-error log, as do the TypeError and ENOTDIR system cases. The old-regex, no-ENOTDIR-log, no-TypeError-log and EDQUOT mutants are red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…art hint copy (#1447 review c) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…n its own (#1440 round-2 review b) One Promise.all over every visible tab let a later transient failure clear a tab already proven Expired, and one slow read hold back another tab's answer. Each run is now asked once, its answer lands when it arrives, and a missing verdict is never taken back. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…row are unreadable (#1434 verification a) The Codex locator skips a file it cannot open, so a listable tree holding the conversation's own unreadable rollout read as absent; assertTreeListable now fails on a file named for the conversation. A Pi file named for the session whose first row is not its session header is damaged, not 'another cwd'. The OpenCode typed query error and the prompt folder's cache-hit parse check are pinned. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… own cwd, and no Cancel on a missing run (#1440 review c) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…; keep a submission recorded before adoption; retry after a reservation collision (#1375 review round 2) Round 2: (a, b) a late Claude/Codex child's readiness waiter could deliver after the parent detached; deliverPromptWhenReady takes a shouldDeliver guard asked when the gate opens, and the late path passes the parent-lease check. (c) the renderer files the child before its create answer reaches main, so a parent's own send could land first and adoption reset the record, causing a second brief; adoption now keeps an existing record. (c) a reservation collision ended the late loop; it now retries. (b) the stop test goes through the real orchestration_send_prompt. All red at ed86637; the send_prompt bookkeeping mutant is red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…e-bootstrap # Conflicts: # src/main/orchestration/OrchestrationBridge.ts
…esolve the parent lease through its alias chain W2's patch (temp/q85): orchestration.bootstrap_mark_failed on the late path only, and the combined late-create + parent-replacement case. isParentAttached now resolves the parent through #1369's currentParentId, so a parent replaced during the late window keeps its brief flowing to the successor; only a parent with no live successor is gone. Removing the alias resolution fails the replaced-parent test. Orchestration suites green; tsc 0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…pts-unreadable # Conflicts: # src/main/conversations/service.system.test.ts
… the fetch timeout and YAML parse detail; pin the file-in-the-way reason (#1456 review round 2 c) Round 2 c: 'Provider failed./var/...' was kept because '.' and '-' counted as word characters; PATH_LIKE now treats a slash after anything but a letter, digit or underscore as a path start. The 60 s fetch abort and the YAML parser branch now log their raw detail, and the file-in-the-way reason has a real-filesystem test. All red at a037caf; the file-component path mutant is red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…eck by identity, not class name (#1456 review round 2 a, b) Round 2 a/b: with prettyErrors off the parser leaves linePos unset, so 'near line N' never appeared; the line now comes from the error's offset and the tests require it. A foreign class merely named GitHubSkillSourceError passed the own-text check; it is now instanceof (githubSkillSource no longer imports the mapper, so no cycle). Their log and FILE_COMPONENT findings were already fixed in ad03125. The old-linePos mutant is red. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…dentity fix(undo-close): a restored agent keeps its derived TLDR identity
…dren fix(undo-close): a restored parent's live children follow it to its new id
…ootstrap fix(orchestration): deliver the bootstrap prompt to a child created after the deadline (#1370)
…ection-fidelity fix(provider-switch): carry the native projection report through switch, duplicate and rewind
…-unreadable fix(conversations): an unreadable conversation file is said, not 'no prompts' (#1306)
fix(workflows): show a run whose stored data is gone as expired, without a Resume that fails
…-start fix(cli-updates): an update that cannot start says so and stays dismissable
…ispose fix(shutdown): dispose the goal loop at quit, after sessions and the MCP host stop
fix(skills): user-visible errors name the problem, never the path or the OS text
Owner
Author
|
Batch disposition (B6): 9 members merged via GitHub into this branch after GATE PASS (member mode) on 53568fb. Their residuals are on each member's disposition and manager verification (#1375, #1434, #1456 manager-verified at the cap). It merges when exact-head CI is green, the body is in past tense and a non-member gate PASS is recorded. |
Owner
Author
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Integration batch P (owner-approved merge mode, owner-requested history form). The branch was created from origin/main 53568fb (batch O); each member PR was retargeted here and merged through GitHub.
Every member had GATE PASS (--member) on main 53568fb, and a local merge probe of all nine was clean. #1450 was held for the next batch (it shares src/main/index.ts with #1449, q125). #1354 was still a draft. Merged after green exact-head CI and a recorded non-member gate PASS (see comments).
🤖 Generated with Claude Code