Use Gemini as a third external reviewer alongside two other vendors: diff review with an APPROVE or REQUEST_CHANGES verdict, and QA-breaker duty in the post-build test ritual. Headless over REST with a CLI fallback, and verdicts go to a shared usage journal. Triggers: "/gemini review", "gemini opinion on this diff".
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Its siblings: Codex (secondop.py / codex_review.py, the default) and Grok
(grok_review.py, a CLI on the hub). Gemini is the third independent vendor: it rescues the
verdict when Codex has burned through its quota, and as a hetero-pair it breaks what the other two
never saw.
cc_review.py / codex_review.py / grok_review.py.The breaker for /tt (the verdict is logged into secondop automatically, no manual log-ext needed):
python "%USERPROFILE%\.claude\scripts\cc-review\gemini_review.py" break --task <task-id> --context "<what we built + what has already been checked>"
Diff review:
python "%USERPROFILE%\.claude\scripts\cc-review\gemini_review.py" review --repo "<path to the repo>"
Is the rail alive (before the ritual leans on it):
python "%USERPROFILE%\.claude\scripts\cc-review\gemini_review.py" doctor
Flags: --range "HEAD~1 HEAD" · --diff <patch> · --task <task.md> (what was asked for) ·
--model <model> · --engine cli|rest · --timeout S · --no-log (break without writing to usage.jsonl).
break: the first line is ACCEPT (agreement) · COUNTER / BLOCK (a finding → that is
/tt Step 4: root cause → fix → re-run, not "an opinion noted for the record"). An unrecognized
verdict → exit 3, the external eye does NOT count (ask again in the right format, or log an
explicit log-skip).review: VERDICT: APPROVE | REQUEST_CHANGES + a report review-gemini-<ts>.md next to the repo.usage.jsonl itself and prints ⚠️: the
/tt verdict is then ⚠️ PARTIAL at best, never ✅. A missed call ≠ a green test.GenerateRequestsPerDayPerProjectPerModel-FreeTier =
20 requests per day per model per project, and there is ONE project for the whole fleet
(laptop + hub + anchor node share the bucket). The model chain gives ~4 buckets a day; Pro models
on the free tier return 429 immediately. Going over = a 429, not a bill (billing is not attached
to the project). When the window is burned the script waits exactly as long as Google tells it to
and retries EXACTLY once (GEMINI_NO_RETRY=1 disables that). Need more — either a second key from
a DIFFERENT Google account (the bucket is per project), or billing on the project (money → the
operator decides).IneligibleTierError / UNSUPPORTED_CLIENT,
"migrate to Antigravity"). Don't try to log in again — only the API key works
(~/.gemini/settings.json → security.auth.selectedType: "gemini-api-key").--engine cli is slow (measured 2026-07-27: minutes versus ~10s for REST) — kept as a second
rail in case REST starts erroring; the default is rest.npm i -g @google/gemini-cli (only needed for --engine cli),
gemini.env arrives through the secrets-store sync, ~/.gemini/settings.json = auth gemini-api-key.Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.
Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).
Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_
Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Run an autonomous consensus round between agent peers on several machines: they negotiate a decision over a shared channel and execute it without a human relaying messages. Use when multiple machines must agree on a change. Triggers: "/03", "work it out among yourselves", "find consensus". Wakes the human only for irreversible actions or a deadlock.
中文简介正在准备中。当前显示原文说明。
Recover a session after a hard crash (app died, machine rebooted, context lost mid-work): replay the turn-state ledger for the last requests, files and decisions, run a green/red health ping over the system map, file sync and MCP connectors, then rebuild the previous session history for pickup. Triggers: "/1", "resume after crash", "wake".
中文简介正在准备中。当前显示原文说明。
Show the operator's day: today's calendar plus what needs them personally (leads due, people awaiting a reply, deadlines), pulled from Calendar and Telegram and ranked. Read-only: sends and commits nothing. Triggers: "/agenda", "what's on today", "my agenda".
中文简介正在准备中。当前显示原文说明。
Rewrite machine-written text so it reads like a busy human wrote it: cut 50-70%, break the even rhythm, drop AI-tell words and filler, keep facts and protected terms. Use before publishing any post, reply or email drafted by a model. Triggers: /ai-slop, /slop, humanize, anti-slop, make it shorter and alive.
中文简介正在准备中。当前显示原文说明。
Decision protocol for strategic work: recall the knowledge vault, run a gap analysis, emit a deep-research prompt for external tools, then synthesize the returned reports into a decision memo. Use before any new product, feature, market, pricing, GTM or architecture call, where local recall alone is not enough. Triggers: "/alfa-search-recall-deepresearch", "alpha protocol", "R+DR".
中文简介正在准备中。当前显示原文说明。
- КРЕДИТ АВТОРУ ИДЕИ: совет из комментария (TG @ClawRus / @ClawInga, комменты под постами Антона в FB, треды и issue на… Триггеры: “/alpha-credit“, “/credit“, “кто нам это посоветовал“, “запиши совет“, “поблагодари советчиков“, “кому мы должны спасибо“, “кредит автору“, “чей это был совет“, “credit the author“
中文简介正在准备中。当前显示原文说明。