feat(voice-prompt): 보이스 모드 STT 입력 정규화 플러그인 — 3단 분류 + 저장소 대조 식별자 해소 #193
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,5 @@ | ||
| { | ||
| "name": "voice-prompt", | ||
| "version": "0.1.0", | ||
| "description": "Normalize Korean voice-mode STT input before acting on it. Strips speech fillers, fixes orthography and Korean-to-English code-switching, resolves garbled identifiers (file, function, branch, skill names) against the actual repo instead of guessing, and never rewrites numbers or dates. Echoes one line of what it understood, then proceeds; gates only irreversible actions. Sibling of plaud-note-taking, which corrects archival transcripts instead of live commands." | ||
| } |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,10 @@ | ||
| { | ||
| "name": "voice-prompt", | ||
| "version": "0.1.0", | ||
| "description": "Normalize Korean voice-mode STT input before acting on it. Strips speech fillers, fixes orthography and Korean-to-English code-switching, resolves garbled identifiers (file, function, branch, skill names) against the actual repo instead of guessing, and never rewrites numbers or dates. Echoes one line of what it understood, then proceeds; gates only irreversible actions. Sibling of plaud-note-taking, which corrects archival transcripts instead of live commands.", | ||
| "author": { | ||
| "name": "YoungjaeDev" | ||
| }, | ||
| "license": "MIT", | ||
| "skills": "./skills/" | ||
| } |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,63 @@ | ||
| # voice-prompt | ||
|
|
||
| One skill, `voice-prompt`, that normalizes Korean voice-mode STT input before the session acts | ||
| on it. Typed once, it stays active until released. | ||
|
|
||
|
Comment on lines
+1
to
+5
There was a problem hiding this comment. 📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
영향: Claude 지침이 As per coding guidelines, Also applies to: 59-68 🤖 Prompt for AI AgentsSource: Coding guidelines |
||
| ## Why this exists | ||
|
|
||
| Korean voice mode breaks in three ways: orthography misrecognition, failed Korean-to-English | ||
| code-switching (an English identifier spoken with Korean pronunciation comes back as Korean | ||
| syllables — "로더 파일"), and speech habits landing inside the command ("어쨌든", "그냥", "뭐"). | ||
|
|
||
| Two of the three barely need a plugin. A model already ignores fillers and reads through | ||
| misspellings, so making that explicit buys consistency, not new capability. The third one is | ||
| different in kind: hearing "로더 파일", a model **guesses** a filename. It cannot know whether | ||
| `loader.py` exists without looking. That is not a comprehension failure, it is a missing action — | ||
| and a missing action is exactly what instructions can add. | ||
|
|
||
| So this plugin is not a text cleaner. It is a threshold changer. Everything it really | ||
| contributes is a decision rather than a transform: | ||
|
|
||
| | What it adds | Kind | | ||
| |---|---| | ||
| | Resolve identifiers against the repo instead of guessing | action added | | ||
| | Never rewrite numbers, dates, versions, PR/issue numbers | action forbidden | | ||
| | Echo one line of what was understood before acting | observability | | ||
| | Confirm before an irreversible action | gate | | ||
|
|
||
| ## Shape | ||
|
|
||
| ``` | ||
| plugins/voice-prompt/ | ||
| ├── CLAUDE.md | ||
| ├── .claude-plugin/plugin.json | ||
| └── skills/voice-prompt/ | ||
| ├── SKILL.md # the body | ||
| ├── references/stt-error-classes.md # error taxonomy, live-command stance | ||
| ├── references/korean-filler.md # filler classes + function-residue test | ||
| └── templates/speech-profile.md # seeded to .claude/voice-prompt/speech-profile.md | ||
| ``` | ||
|
|
||
| ## Boundary with plaud-note-taking | ||
|
|
||
| Both correct STT output, and the error distribution is the same, so the taxonomy is shared in | ||
| spirit. The discipline is not. `plaud-note-taking` treats its transcript as immutable evidence | ||
| and preserves every filler because the filler is part of the record; this skill treats its input | ||
| as a disposable command and deletes the filler because the filler is noise. Their term | ||
| dictionaries do not overlap either — one holds meeting attendees and company names, the other | ||
| holds one speaker's pronunciation habits. | ||
|
|
||
| Two plugins, not one skill with a mode flag: a body carrying both stances would need a | ||
| conditional on every rule. | ||
|
|
||
| ## Known limit | ||
|
|
||
| This is instruction-following, not interception. Nothing sits between STT and the model — a | ||
| skill cannot rewrite the incoming message, and `UserPromptSubmit` hook output arrives as | ||
| appended `additionalContext` rather than a replacement. So the normalization happens in the | ||
| model's own reasoning and carries the same reliability as any other skill: it can fade in a long | ||
| session. | ||
|
|
||
| The echo line is the tell. If it stops appearing, the skill has drifted out — re-invoke | ||
| `/voice-prompt`. That self-signal is why the echo is mandatory rather than optional, and why | ||
| this plugin ships no marker file or re-injection hook to enforce stickiness. | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
For a marketplace or local plugin install, this skill is exposed under the plugin namespace, consistent with the repository's existing
/github-dev:resolve-issueexample, so/voice-promptdoes not identify the added skill. Because automatic activation is explicitly disabled, users following this primary instruction cannot enable the feature; document/voice-prompt:voice-promptconsistently here, in the skill description, and in the plugin guide.Useful? React with 👍 / 👎.