Menu
Text-to-speech via Microsoft Edge TTS (free, neural voices).
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
ABCd context protocol across ABC root files (AGENTS.md, BACKLOG.md, CHANGELOG.md), human entrypoint README.md files tree, and /docs. Use after meaningful project changes, backlog drift, documentation refactors, self-evolving context maintenance, or a forced context reconciliation pass.
Build, refactor, review, or debug Svelte 5 components that use Bits UI primitives. Use when working with bits-ui dialogs, popovers, dropdowns, comboboxes, selects, tabs, date/time controls, menus, tooltips, portals, render delegation, or Bits UI type helpers.
Collaborative idea-to-design and inquiry protocol. Use for product/architecture exploration, research-style question shaping, feature design, standards, specs, UX concepts, module boundaries, and non-trivial behavior changes when uncertainty matters.
Pragmatic senior-engineer operating contract for coding agents. Use for coding, implementation, refactoring, debugging, code review, tests, validation, repository maintenance, file edits, or technical investigation.
Horizontal Gene Transfer protocol for skills. Synchronizes best practices and architectural patterns across the skill library.
Manage a guarded release flow that commits prepared release work on dev, opens a dev-to-main pull request with a release-focused PR summary, waits for checks, merges on success, tags, and optionally publishes an existing npm package. Use when the user asks to prepare or execute a dev→main release PR, hotfix release PR, or Dev2Main PR Summary workflow.
| name | edge-tts |
| description | Text-to-speech via Microsoft Edge TTS (free, neural voices). |
| metadata | {"version":"1.0.18"} |
Direct Node.js client for Microsoft Edge's online TTS service. The canonical client is scripts/say.mjs; scripts/say.sh is the shell entrypoint wrapper. There are no Python, CLI-wrapper, or legacy shell fallbacks.
./scripts/say.sh "Привет" ru +30% # Synthesize and play with an available local MP3 player
./scripts/say.sh "Привет" ru 1 /tmp/hello.mp3 # Save MP3; numeric rate multiplier is supported
./scripts/say.sh "Привет" ru +30% /tmp/a.mp3 /tmp/a.jsonl
./scripts/say.sh --text "Hello" --lang en --write-media /tmp/hello.mp3
./scripts/say.sh --file input.txt --voice ru-RU-SvetlanaNeural --write-media out.mp3 --write-subtitles out.srt
./scripts/say.sh --list-voices
say.sh delegates to say.mjs, which supports both positional and flag-style invocation. Positional form is compact for humans and simple wrappers; flag form is explicit and self-documenting for operators.
--text, --file — text source; - means stdin for files.--lang, --voice — language shortcut or explicit voice.--rate, --volume, --pitch — Edge prosody controls.--boundary — SentenceBoundary or WordBoundary.--write-media — MP3 output path; - means stdout.--write-subtitles — SRT output path; - means stderr.--write-metadata — JSONL boundary metadata output path.--list-voices — direct voice list request.Playback is optional and only applies when no media output path is provided. say.mjs auto-detects a local MP3 player and removes the temp file after playback when the player is blocking.
Default candidates:
ffplay, mpv, vlc/cvlc.mpg123.afplay.Start-Process.If no player exists, generation is still available through --write-media; the agent can then play the MP3 with any system-specific tool. Windows default-app playback uses delayed temp cleanup because registered apps usually detach immediately.
1 = +100%, 0.8 = +80%, -0.5 = -50%.