maestro-frontier
Maestro Frontier local multi-CLI fusion engine - compose or choose a read-only model panel, inspect the catalog, arm, disarm, or run it
- License 许可
- MIT
- 最近更新
- 2026-11-13
- Stars
- ★ 1.6k
- Forks
- 483
01 前言 Front Matter
- name
- maestro-frontier
- license
- MIT
- description
- Maestro Frontier local multi-CLI fusion engine - compose or choose a read-only model panel, inspect the catalog, arm, disarm, or run it
03 SKILL.md 正文
Drive the Maestro Frontier engine: a local multi-CLI fusion engine where a
parallel panel feeds a judge and grounded synthesis. When a user asks to
compose or choose a model panel, route the request to frontier compose after
consulting frontier catalog; never guess a model or preset ID.
When the trusted Maestro Codex plugin hook is installed, arming a non-off
mode makes normal later Codex prompts auto-run through Frontier. run remains
an advanced, manual one-off. Do not edit Frontier state files by hand.
Command launcher
Use maestro when it is on PATH. If it is not, and this skill is loaded from
the Maestro Codex plugin, locate the plugin root by walking up from this
SKILL.md until .codex-plugin/plugin.json is present, then run:
node "<maestro-plugin-root>/bin/maestro.cjs" frontier ...
In the examples below, maestro frontier ... means either the bare command or
that plugin-root launcher.
Catalog and compose a panel
Use project scope from the repository root for state-changing commands:
maestro frontier catalog
maestro frontier catalog --json
maestro frontier compose --models <model>,<model> --dry-run --scope codex-project
maestro frontier compose --models <model>,<model> --judge <model> --synth <model> --effort <level> --scope codex-project
maestro frontier compose --models <model>,<model> --save <name> --scope codex-project
maestro frontier effort <auto|low|medium|high|xhigh|max|ultra> --scope codex-project
frontier catalog is the source of truth for models, presets, aliases,
readiness, and required configuration. compose accepts one to eight
comma-separated models. Judge and synth default to the first panel model.
--dry-run changes nothing; a non-dry run saves and arms the resolved custom
fusion panel. --effort persists one validated level across panel, judge, and
synth stages for selected effort-aware Claude and Codex models. effort auto
returns the armed panel to provider defaults.
Modes, inspection, and saved presets
maestro frontier mode off --scope codex-project
maestro frontier mode single --model <model> --scope codex-project
maestro frontier mode fusion --preset <preset> --scope codex-project
maestro frontier mode fusion --preset custom --models <model>,<model> --scope codex-project
maestro frontier status --scope codex-project
maestro frontier roster
maestro frontier preset save <name> --models <model>,<model> --judge <model> --synth <model> --scope codex-project
maestro frontier preset list --scope codex-project
maestro frontier preset delete <name> --scope codex-project
Use catalog before selecting any model or preset. roster reports only
readiness and configuration names, never secret or configured model values.
Current models, overrides, and release gate
opus pins Claude Opus 5.5 (claude-opus-5-5), fable pins Fable 5.1
(claude-fable-5-1), sonnet-5 keeps Sonnet 5 (claude-sonnet-5), and
haiku adds Haiku 4.5 (claude-haiku-4-5-20251001). Claude Code v2.1.280+
is required for Opus 5.5; v2.1.257+ is required for Fable 5.1.
Opus, Fable, and Sonnet support low, medium, high, xhigh, and max;
Haiku has no effort flag.
Codex selectors include astra (gpt-6-astra), sol (gpt-6-sol),
terra (gpt-5.6-terra), and luna (gpt-6-luna), plus auto-review and
the retained GPT-5.5, GPT-5.4, GPT-5.4 Mini, and GPT-5.3 Codex Spark selectors.
GPT-6 models support low, medium, high, xhigh, and max in Maestro’s
UI, without none or ultra. Earlier explicit selectors and third-party
adapters remain available.
Use frontier catalog for exact aliases and supported effort levels.
MAESTRO_FRONTIER_MODEL_TERRA, MAESTRO_FRONTIER_MODEL_LUNA, and
MAESTRO_FRONTIER_MODEL_SOL remain optional model-ID overrides. Codex Desktop
/ IDE sessions read those overrides from ~/.codex/.env; restart the app and
open a new thread after changing it.
Before releasing Codex catalog changes, probe every first-party Codex selector:
node frontier/smoke.cjs
This explicit release gate invokes those selectors through the normal read-only dispatch path.
Normal and debug use
After mode is non-off, use ordinary Codex prompts; the trusted hook supplies
the Frontier result. Turn it off or run one manual prompt with:
maestro frontier mode off --scope codex-project
maestro frontier run "<prompt>" --scope codex-project
All panel, judge, and synthesizer subprocesses are one-shot and read-only:
they return text only and never edit the workspace, commit, or run autonomous
loops. On ERROR [<reason>]: <detail>, relay the reason.
Indicator instruction
<!-- INDICATOR INSTRUCTION - apply every turn while this skill is active -->
Before composing your first reply each turn, run:
maestro frontier status --scope codex-project
Parse the returned JSON mode field:
- If
modeis notoff, make the first line of your reply exactly:Maestro Frontier ON (<label>)where<label>issingle - <model>orfusion - <preset>; for a custom preset usefusion - custom (<model1>, <model2>, ...). - If
modeisoff, output no indicator line.
<!-- END INDICATOR INSTRUCTION -->
04 同仓库更多 Skills
migrate-to-e2e
Moves an existing Playwright, Cypress, Selenium, Appium, Detox, Maestro, or TesterArmy suite to e2e one slice at a time, with the old suite running beside the new one until each ported test is proven. Keeps a MIGRATION.md progress file so any later session resumes where the last stopped. Optionally plants a defect to show both suites catch it. Use for /migrate-to-e2e, "migrate our tests to e2e", "port these Maestro flows", "move our TesterArmy tests", or when a project has an e2e config and an older suite still running.
e2e
Writes, runs, and debugs end-to-end tests for web and mobile apps with e2e, the agentic test runner, which mixes agent.act goals with exact locator assertions on a Playwright browser or an iOS or Android device. Also drives a project's running app live over the e2e MCP server to look at a screen, check a UI change, take a screenshot or record a video, and try a locator; and runs bug bashes, parallel e2e explore runs with each bug proven by a repro test. Use when a project depends on e2e or has an e2e.config.ts, when asked for end-to-end, browser, mobile, or agentic UI tests, to check a change in the running app, to QA or bug bash an app or branch, or when an e2e run fails.
create-verification-skill
Generate a project-local verify-<app> skill on top of e2e, so any coding agent can launch the app, check it is worth driving, exercise a feature the way a user does, keep evidence, and bug bash it. Writes the skill, a feature map with one explore charter per feature, and proves it once. Use for /create-verification-skill, "make a verification skill for this repo", or when a project has no scripted way to prove UI behavior.
deja-search
deja-vu memory — search the user's past AI coding sessions with the deja CLI. Holds what they already settled and never wrote into the repo (naming rules, banned or already-tried tools, required checks before merge, preferred values, who owns what). Use before your first edit in a task, before you call a change done or ready to merge, before you recommend a tool, library, design or reviewer, before re-debugging an error that may already be solved, when they say 'didn't we fix this before', 'what did we decide about X' or 'I already have that', and before telling them something on this machine does not exist.
deja-reflect
End-of-session reflection and memory curation with deja-vu. Review what was decided, what commands broke and what repaired them, and record durable project rules or decisions with deja remember so the next session starts with them.
revid
Make, edit, export and post AI videos with Revid (revid.ai) through the Revid MCP server. Use whenever the user wants a short-form or long-form video made (from an idea, a script, a URL, an article, a PDF, a tweet, a product page, their own clips, or a song), wants to find what is going viral on TikTok, YouTube, Instagram or X before making something, wants to edit or re-export an existing Revid project, generate a standalone AI image or clip, make a series episode, or publish and schedule a finished video to connected social accounts. Covers setup (hosted MCP with OAuth, or an API key), the 36 tools, credits, polling and the rules for publishing. Routes to the more specific revid-* skills when one fits.