Which agents work with it?

Any agent that reads skills from a directory. Claude Code, Codex and Gemini CLI read it natively; Cursor and VS Code connect over MCP.

The voice runs on your own machine, so there is nothing to meter. The first run downloads it, about 340 MB.

Yes. The skill is MIT-licensed and the authoring format is public.

Create

gstack’s /autoplan now runs a CEO review that asks you nothing

1 hour ago

@gstack Subscribe

gstack's /autoplan --ceo-only runs the CEO review of your plan on its own and asks you nothing. Also: /ship sets review depth by risk instead of diff size, /retro shows friction by skill and keeps its history across worktrees, fixes for Claude Code sessions and brain-sync, and new support tiers for Windows and Gemini. Covers gstack 1.92.23.0 to 1.92.77.0 (55 releases, 2026-11-01 to 2026-11-02), for Claude Code users. Changelog: https://github.com/garrytan/gstack/blob/main/CHANGELOG.md Support tiers: https://github.com/garrytan/gstack/blob/main/docs/support-tiers.md CEO-only runs: https://github.com/garrytan/gstack/blob/main/docs/unattended.md gstack: https://github.com/garrytan/gstack

Ask about this presentation

Show transcript

With /autoplan --ceo-only, gstack’s CEO review of your plan runs to the end on its own. Each question takes its recommendation. It approves nothing, and challenges to your plan wait for you.
Slide text

gstack’s /autoplan now runs a CEO review that asks you nothing

/autoplan --ceo-only
GSTACK_RESULT: skill=autoplan status=reviewed

questions each one takes its recommendation, logged in ceo-review.md
your call it approves nothing; challenges to your plan are listed at the top

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.50.0 · 1.92.58.0
/ship now saves the full reviewer panel for code that touches security, money, auth or secrets. Everything else gets a light review, however big. After two rounds, an open blocker stops for you.
Slide text

/ship now sets review depth by risk, at any diff size

full panel security, user data, money, secrets, auth, merge checks
light other code: one batched reader plus the adversarial pass
one reader docs, tests, eval write-ups
round 2 an open blocker stops for you; there is no round 3

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.45.0
/retro now lists the three skills with the most questions, errors and aborts. In a Claude Code worktree, it now compares against last week.
Slide text

/retro now names the skills that fought you this week

Friction by skill
/ship 7 in 12 runs (4 questions, 3 aborts)

Questions count only when question tuning is on.

in a worktree First retro recorded compared with last week

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.26.0 · 1.92.76.0
/autoplan no longer locks a Claude Code session after an init that changes folders. The timeline stop hook returns within its two-second budget. And overlapping profile updates stop dropping /office-hours sessions.
Slide text

Claude Code sessions stop locking up, stalling and losing office hours

/autoplan no longer locks the session after an init that changes folders1.92.50.0
The timeline Stop hook returns within 2 s when the host holds stdin open1.92.49.0
Overlapping profile updates keep every /office-hours session /plan-tune edits still skip the lock · 1.92.69.0

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.50.0 · 1.92.49.0 · 1.92.69.0
brain-sync used to stall for good when two machines edited one file. Now it keeps theirs and stashes yours. A profile with conflict markers stays local until you repair it.
Slide text

brain-sync stops sticking at push_failed and holds back a broken profile

before push_failed on every run

now their version kept · your edit stashed

bad JSON held as invalid-json until repaired

gstack-developer-profile --profile now exits 1 on a broken file.

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.38.0 · 1.92.46.0
gstack gives full support to macOS and Linux, with Claude and OpenAI models. Windows and Gemini get install and the core skills. Other bugs there close as won't fix.
Slide text

gstack now gives Windows and Gemini minimal support

full macOS and Linux, Claude and OpenAI models
minimal Windows and Gemini: install, /ship, /review, /qa-only, /investigate, /browse

Other Windows-only and Gemini-only reports close as won’t fix.

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.43.0
A local main behind origin no longer turns /ship into a full review. An unfinished cherry-pick stops it by name. A huge pull request carries its receipt in seconds. I'd upgrade for the risk-based review alone.
Slide text

/ship fixes for a stale main, a half-done merge and huge PRs

A local main behind origin no longer forces a full review one run cost about 15–20M tokens · 1.92.76.0
An unfinished cherry-pick or merge stops /ship by name 1.92.25.0
A PR with thousands of files carries its receipt in seconds was about 3 hours per base move · 1.92.42.0

CHANGELOG.md · docs/support-tiers.md garrytan/gstack

CEO review Risk Retro Sessions Sync Windows Fixes
garrytan/gstack · CHANGELOG.md · 1.92.76.0 · 1.92.25.0 · 1.92.42.0