github garrytan/gstack v1.101.0.0
gstack v1.101.0.0

one day ago

gstack v1.101.0.0

What's new

  • Before you're asked to try an app build, the new gstack-first-use command has an agent from another model family play that build on the same kind of device and check the first five things you'll do with it. (#4465)
  • Before an agent asks its owner a question, gstack-owner-answers can check the draft against the owner's earlier answers and suggest dropping only the lines an existing standing yes already covers. (#4467)
  • GitHub release notes now list every change in plain words, with a link to its pull request. (#4390)
  • When another agent drives /ship step by step, each phase now runs as one command, the calls are counted, and the run stops itself after 100 minutes. (#4451)
  • /ship's push step now always ends with one PUSH: line that says whether it pushed; if a script matched the old PUSH: lines, update it. (#4474)
  • /ship's review plan now sends Swift code that asks for privacy permissions or touches health, calendar, contacts or on-device files to the full review panel. (#4472)
  • gstack-autoplan revise --plan with a different file now says the run follows that file. (#4470)
  • /office-hours explains why it recommends a mode, quoting your own words. (#4450)
  • /office-hours keeps its 34 founder videos and essays out of the skill file, so the skill is about 8.6 KB smaller. (#4463)
  • Unattended /autoplan now runs the second-model review you turned on, and waits for it instead of ending the session. (#4459)

Security

  • A security finding's file, line and reason no longer appear in a pull request description; it shows an opaque id and the details stay in the private approval record. (#4486)
  • The secrets check now catches a secret built from short pieces with + or written in a comment, and stops flagging three kinds of ordinary code. (#4457)

Fixes

  • /ship with a token cap on Claude Code no longer stops at its first handoff: the new session now receives its prompt. Thanks @exGeni. (#4452)
  • On Claude Code, that handoff session's prompt now starts with the typed /ship command, so the skill actually loads. (#4487)
  • /ship no longer stops for an owner's sign-off when merging the base only changed base-side text. (#4458)
  • A reviewer's confirming prose no longer becomes a Critical finding that only an owner can clear. (#4458)
  • When /ship stops at its review-round limit and the owner allows one more round, that round now pays for fixing the findings instead of another read-only review. (#4479)
  • /ship on Codex no longer stops at the test coverage audit: Codex skills now start their audit and reviewer subagents with Codex's own tools (turn on multi_agent = true under [features] in ~/.codex/config.toml). (#4489)
  • /ship no longer reports zero blockers when the second-model review found one, and won't finish while that blocker is open. (#4485)
  • When /ship posts a pull request description, the count of earlier review rounds now always comes from the PR's real history; a description carrying extra history blocks could make gstack count one round where there were three. (#4440)
  • gstack-post no longer flags /ship's own run ids and commit ids as phone numbers, so posting a /ship pull request description no longer needs a confirm. (#4488)
  • /ship's review plan now sorts Swift and Xcode files by risk instead of sending every iOS pull request to the full review panel. (#4469)
  • /ship's pre-push test selection now finds the tests of a monorepo app that imports through its own @/ alias. Thanks @dtoelke. (#4464)
  • A test, or a scratch repository in the temp directory, can no longer write approval rows into your real Capy Drive. (#4466)
  • gstack's router no longer gives bug reports a special rule, so the model fixes a bug the way it would without gstack; on Claude Code, a bug report also no longer loads the router. (#4442)
  • An unattended /autoplan run on Claude Code no longer reports its phase guard as off while it is on, and a phase report written only in reasoning now gets a denial that names the fix. (#4481)
  • gstack-autoplan revise no longer refuses a plan that the first next accepted. (#4470)
  • A second-model plan review in /autoplan is no longer reported as unavailable when it actually ran. (#4471)
  • Unattended /autoplan now records the owner's relayed answer to its final question instead of refusing it. (#4484)
  • Two /autoplan runs started in the same second no longer share a run id. (#4459)
  • Two gstack-owner-answers add runs for the same answer no longer collide. (#4490)
  • skill_prefix=true now survives updates when gbrain is installed. Thanks @fliaron. (#4454)
  • The OpenClaw docs no longer send you to ClawHub, which has no current gstack skills; they point to ./setup --host openclaw. Thanks @exGeni. (#4454)
  • gstack-issue-links no longer reads other repositories' issues as gstack's. (#4454)
  • /sync-gbrain no longer says OK when pages failed to import or an artifacts push failed, and one malformed queued path no longer stops artifacts sync for every project. (#4455)
  • Packing a checkout for a Ubicloud run no longer fails when git is rewriting a ref beside it. (#4447)
  • With a custom Codex provider, the Codex readiness check and the second-model review now pick the same model: the provider's model in ~/.codex/config.toml counts only when it is the review tier's model (gpt-6-astra for the frontier tier), and both refuse any other name with the same reason. On the stock OpenAI provider, the second-model review in /autoplan and /spec now uses the review tier's model instead of config.toml's. (#4491)
  • gstack-artifact validate and ack now refuse an /autoplan run.json that gstack-autoplan didn't write, or whose journal, review outputs or checkpoint disagree with it. (#4494)

Changes for maintainers

  • A CI run that passed only after its failed jobs were re-run is now marked as such wherever gstack reuses it. (#4473)
  • /ship reuses CI's passing test run when merging the base changed no file, and stops instead of quietly starting hour-long local test runs or reviewing an empty diff. (#4461)
  • A preregistered plan for testing /ship without its separate coordinating subagent. (#4456)
  • That plan now lets case reverts that conflict with later code be resolved by hand. (#4460)
  • CI uploads free-suite test results to Trunk with analytics-uploader v2.1.8 and analytics-cli 0.17.0. (#4453)
  • A policy-floor test no longer fails once gstack's own version reaches the minor release after the newest policy key. (#4495)
  • The /plan-ceo-review HOLD SCOPE eval no longer fails a review because its follow-up question didn't repeat the mode name. (#4447)

Compare v1.100.0.0...v1.101.0.0 · Full changelog

Don't miss a new gstack release

NewReleases is sending notifications on new releases.