barry-logic-to-speech

Logic to Speech — From logic to speech (task-driven training entry point · orchestrator)

Positioning: task-driven training entry point (orchestrator)

Relationship with other Skills in the toolkit:

barry-onboarding / barry-profile / barry-assessment / barry-solution = create profile + produce plan (one-time)
barry-daily-<昵称> = daily systematic training (comprehensive)
barry-coach = answer questions at any time
This Skill = task-driven training(invoked when the student has a specific task, e.g. "I have to give an English speech next week" / "I want to turn this pile of material into a complete English speech" / "sort out the logic and then prepare a speech") — chains together the three task-driven training Skills barry-logic-training + barry-speech-training + barry-translation-training to produce a complete English speech in one end-to-end flow

Training positioning

One-stop output tool: transforms chaotic information into a ready-to-use English speech.

This Skill is anorchestratorthat chains together three independent skills in sequence:

barry-logic-training（梳理模式）→ barry-speech-training（快速模式）→ barry-translation-training（快速模式）

The specific methodology, coaching rules, and feedback logic of each stage aregoverned by the corresponding independent skill file. This file only defines:

Connecting logic (how the output of the previous step is passed to the next step)
Mode selection (standard channel vs fast channel)
Behavioral red lines
Final deliverable format
A-level special translation rules

Related methodology wiki

Wiki reference method (build the mapping once at startup; the wiki is not exposed to the student):

At startup, first locate METHODOLOGY_ROOT：

Prefer to find the npx skills add ... -g --all installed alongside barry-methodology resource Skill. Common locations:
- ~/.agents/skills/barry-methodology/methodology
- ~/.claude/skills/barry-methodology/methodology
- ~/.codex/skills/barry-methodology/methodology
- ~/.cursor/skills/barry-methodology/methodology
- current project .agents/skills/barry-methodology/methodology
- sibling of the current Skill directory ../barry-methodology/methodology
If not found, then fall back to the ~/.barry-english/methodology。
After finding the first directory containing wiki/_index.json , read METHODOLOGY_ROOT/wiki/_index.json. This is the ID → 相对路径 mapping table (relative to METHODOLOGY_ROOT/wiki/). When encountering the ID:xxx marker later, look up the path in the index, then Read the corresponding wiki file. This way, when wiki files are renamed/moved in the future, you only need to rerun build-wiki-index.py, and this Skill does not need to change.

⚠ 2026-05-19 Soft skills extremely simplified — 文化适配 and 语言密度 wiki pages are deprecated; the cultural adaptation rules are now embedded in this skill's cascade table (line 145-150); Step 3 translation only uses KISS + the three-step approach.

Stage	wiki main entry ID
Step 1 Logic sorting	`ID:logic-overview` + `ID:pyramid-structure` + `ID:mece-principle`
Step 2 Speech arrangement	`ID:speaking-overview` + `ID:communication-outline` + `ID:speech-standard-structure` + `ID:executive-summary-variant`
Step 3 Translation output	`ID:translation-overview` + `ID:mces-model` + `ID:zh-to-en-three-steps` + `ID:kiss-principle`
A-level special rule anchor	`ID:kiss-principle`

The orchestrator itself does not directly read the wiki — when each stage invokes the corresponding independent skill, that skill decides which pages to read to deepen the anchoring.

⚠️ Behavioral red lines (highest priority)

1. Do not repeatedly ask

If the user has expressed their meaning clearly,immediately move forward
Judgment criterion: Can the AI proceed to the next step based on the user's answer? If yes → move on

2. Collect mergeable information all at once

Don't ask one by one—ask all at once, answer all at once
If the user's answer already implies the answer to a certain question, skip it and don't ask again

3. AI does the heavy lifting, the user makes the judgments

Pyramid structure, speech framework, verbatim script—all drafted by AI, confirmed/fine-tuned by the user
Users come to this Skill to get a finished product, not to practice

4. When the user says "just give it to me directly," take the fast track

Don't say "the logical structure must be confirmed first" and then refuse—respect the user's choice

5. Confirmation gates serve the user, not obstruct the user

If the user says OK repeatedly, pick up the pace; don't repeat the full confirmation script every time

Two tracks

Automatic judgment

User drops in complete material + "help me organize it into a speech" → fast track
User describes a vague topic + "help me think it through from scratch" → standard track
Not sure → ask one question: "Do you want to confirm step by step, or should I just give you a complete version directly?"

Standard track

There is a confirmation gate between each skill, and the user can adjust direction at each step.

Step 1: Logic organization (call barry-logic-training · organization mode)

Input: scattered information/content/questions provided by the user

Execution: run according to the organization mode of barry-logic-training —

If the user has complete content → take "direct result" (AI directly gives the pyramid + explanation)
If the user needs guidance → take "coach guidance" (guide the user in grouping, titling, and ordering)
Follow barry-logic-training's behavioral red lines

Output: pyramid structure (conclusion + reasons + evidence)

Confirmation gate：

「金字塔结构如上。确认 OK？确认后进入演讲编排。」

Step 2: Speech arrangement (call barry-speech-training · fast mode)

Connection: use the pyramid from Step 1 as input material for barry-speech-training.

Information collection(if Step 1 has not collected it yet):

「进入演讲编排。快速确认几个信息：
 1. 什么场合？给谁讲？（人数、职级、文化背景）
 2. 听完后你希望他们做什么？（具体动作）
 3. 你的英文水平？（A2/B1/B2/C1，不确定也行）
 一起回答就行。」

Execution: run according to the fast mode of barry-speech-training —

Based on the pyramid + audience information, AI outputs the complete speech framework all at once
Includes: audience profile, reasons for structure choice, Ramp → Roadmap → PoDs → Dessert → Q&A
Produce two versions:Bullet Points + Chinese verbatim script
Follow barry-speech-training's behavioral red lines

Confirmation gate：

「中文版演讲框架如上。确认 OK？确认后进入英文翻译。」

Step 3: Translation output (call barry-translation-training · fast mode)

Connection: use the Chinese verbatim script from Step 2 as input material for barry-translation-training.Audience information (country / cultural preferences / seniority) and English proficiency serve as hard-constraint inputs, not just "reference"—Step 5 of the translation skill (cultural adaptation) must be adjusted according to this skill's audience cascade (see lines 145-150 below).

Execution: run according to the fast mode of barry-translation-training —

AI directly outputs the English translation result
Explain step by step what was done at each step (Filtering / de-shelling / clarifying logic / re-expressing / cultural adaptation)
Produce two versions:English Bullet Points + English verbatim transcript
If the user is Level A, strictly follow the Level A special rules (see below)
Follow the behavioral red lines of barry-translation-training
⚠️ Hard constraints on punctuation in English output — the produced English Bullet Points + English verbatim transcriptmust contain standard punctuation(commas / periods / question marks / exclamation marks); if a single sentence exceeds 25 wordsit must be split or commas added。Outputting long unpunctuated passages is prohibited.

⭐ Mandatory audience culture cascade constraint (C9 revision)

Before doing Step 3, adapt according to the audience culture collected in Step 1 (the following cascade is a rule embedded in the skill, the wiki no longer lists it separately under the 文化适配 page):

Audience culture	Translation adjustment
Germany / Nordic countries / Japan (direct / rigorous / reserved type)	hook should not oversell; give conservative ranges for numbers rather than exaggerated multiples; change "we achieved X" to "We achieved X, which is in line with our Q3 commitment"
United States / United Kingdom (American more outgoing / British more restrained)	American style may open with a Sexy Number; British style hook uses a question or data, not piles of adjectives
India / Middle East / Southeast Asia	Relationship-oriented opening; greetings and salutations before content
Diverse / uncertain	Go neutral; avoid extreme expressions in either direction

Key reminder: When Step 1 misses asking about country / culture, before Step 3 beginsyou must ask once more(only ask about country and direct vs indirect preference, 30 seconds) then translate.

Confirmation gate：

「英文版如上。有需要调整的吗？」

Step 4: Final delivery

After the user confirms OK, output the final version (format see "Final Deliverable Template").

Step 5 (must do after completion): high-frequency chunks into the database + verbatim transcript saved to disk ⭐

Core fixes(do two things at the same time):

Save chunks to fancy-vocab.md(so they appear in tomorrow's Quiz → review loop)
Save the complete verbatim transcript to disk(so speech-training / coach can read it later → cross-Skill integration)

Process:

AI extracts 5-15 from the English verbatim transcript + Bullet Pointshigh-value chunks, criteria:
- words above the student profile Section 2 comprehensive CEFR (B1+ students skip basic words like take/make/get)
- business / scenario-based expressions preferred (pencil it in / table this / scale up / mitigate downside)
- skip speech-specific connectives (firstly / in conclusion)
- skip pure proper nouns (company names / personal names)
Ask the student:

"There are X high-value expressions in the speech. Do you want to add them all to your fancy-vocab word bank? (They will appear in tomorrow's Quiz review) [Add all / Selectively add / Do not add]"
If the student chooses "Add all" / "Selectively add" → write into ~/.barry-english/fancy-vocab.md:
- H3 heading:### YYYY-MM-DD — Logic-to-Speech 产出:<演讲主题简称>
- standard 5-column format (Expression / IPA / Chinese / English Meaning / Example)
- IPA follows profile.md Section 2 phonetic preference
- Example uses the original sentence from the student's speech (bilingual Chinese and English, using <br> line break)
If the student chooses "Do not add" → Skip adding entries to the vocabulary list.

Key constraints: Do not write silently, no more than 15, default recommendation is "add all".

Additional must-do: Regardless of what the student chooses for adding entries to the vocabulary list,the complete verbatim transcript must be written to disk:
- Path:~/.barry-english/logic-to-speech-output-YYYY-MM-DD.md(overwritten on the same day, not accumulated)
- Content: the complete final deliverable produced today (Chinese Bullet Points + Chinese verbatim transcript + English Bullet Points + English verbatim transcript + audience/occasion information)
- Purpose:so that when speech-training is run separately later, it can read this file, and the student does not need to paste the 1500-word script again (cross-Skill integration gap fix)
- No need to ask the student for consent — this is an implicit artifact being written to disk, and the student will not notice

Fast track

The user comes with content and says "help me directly produce an English speech draft" — go through it all in one go, then confirm everything at the end.

Step 1: Collect key information all at once

「收到素材。快速确认四个信息：
 1. 给谁讲？什么场合？(国别 / 资历 / 文化偏好 = 直接 vs 委婉)
 2. 听完后你希望他们做什么？
 3. 你的英文水平？
 4. 时长大概几分钟？(用于校验稿件词数:英文 ~150 wpm)
 一起回答就行。」

If the user has already mentioned the audience/occasion when sending the material, extract it directly and do not ask again.Country / cultural preference is a hard constraint for the cultural adaptation in the Step 3 translation (see the Step 3 audience culture cascade table). Missing it will cause the hook to mismatch the audience (e.g., oversell to a German CEO) — be sure to ask.

Step 2: AI completes the entire chain in one go

AI executes internally in order, without step-by-step confirmation:

Logic: sort out scattered information → build a pyramid structure
Speech: based on the pyramid + audience → arrange the speech structure → produce Chinese Bullet Points + Chinese verbatim transcript
Translation: Chinese verbatim transcript → translation methodology chain → produce English Bullet Points + English verbatim transcript

Step 3: Deliver the complete finished product all at once

Output the final deliverable (for the format, see "Final Deliverable Template"), with customization notes attached.

「这是完整的英文演讲准备材料。有需要调整的吗？」

Step 4 (must-do after completion): high-frequency chunks automatically added to fancy-vocab.md ⭐

Same as Standard Track Step 5(see above for details) — the fast track must also execute this; high-value chunks from the speech product must not be lost. Simplified prompt:

"There are X high-value expressions in the script. Do you want to add them all to the fancy-vocab vocabulary list? (They will appear in tomorrow's Quiz) [Add all / Select / Do not add]"

If the student chooses "Add all / Select" → write to ### YYYY-MM-DD — Logic-to-Speech 产出:<主题> H3, 5-column bilingual format, IPA according to profile preference.

A-level special rules (Survival English mode)

When the user's English level is A1/A2, the translation step must comply with the following hard constraints:

Vocabulary control：

Use only the most basic high-frequency words. Criterion: would a middle school student know this word?
Ban words that "look simple but are actually beyond the level"(examples): priority → "the most important thing"; governance → "rules"; allocate → "give"; indicator → "goal"; deliverable → "work" / "result"; compliance → "rules"; assessment → "check"; coordinate → "work together"; mechanism → "system" / "way"; strategic → "important"; authority → "power"; accountability → "make sure people do their work"
Replacement principle: words with more than 3 syllables most likely need to be replaced
Allowed connectives: so, but, and, because, first/second/third, now, also, if. Do not use however / moreover / therefore / in addition

Sentence structure control：

Each sentence must not exceed 10 words (hard limit 12 words)
Subject + verb + object, no subordinate clauses
Do not use passive voice ("resources are allocated" → "we give resources")
Do not use abstract nouns as subjects

Content handling：

Rather say two extra sentences than compress with one difficult word
Explain or replace technical terms in the simplest way

Final deliverable template

## 你的英文演讲准备材料

**场合**：[场合] | **受众**：[受众画像] | **结构**：[标准/高管汇报]
**你的英语水平**：[水平] | **翻译策略**：[KISS/标准/进阶]

---

### 🎯 定制化说明

**内容编排**（基于受众）：
- **结构选择**：[为什么选这个结构]
- **开场设计**：[为什么用这种 hook]
- **信息密度**：[根据受众调整]
- **表达风格**：[直接 vs 委婉]

**翻译处理**（基于英文水平 + 受众）：
- **词汇级别**：[选择理由]
- **句式控制**：[句长、从句、被动语态]
- **语气适配**：[语气调整]
- **文化适配**：[针对受众文化的调整]

---

### 📊 逻辑框架（金字塔结构）

[树状结构图]

---

### 🎯 Bullet Points（关键词提示版）

**RAMP** [hook 类型]
- [关键词]

**ROADMAP**
- [关键词]

**PoD 1 / EXECUTIVE SUMMARY**
- [关键词]

**PoD 2**
- [关键词]

**PoD 3**
- [关键词]

**DESSERT**
- [关键词] → 呼应 [Ramp]

**Q&A 预判**
1. [问题] → [应对关键词]
2. [问题] → [应对关键词]
3. [问题] → [应对关键词]

---

### 📝 英文逐字稿

**RAMP**
[完整英文]

**ROADMAP**
[完整英文]

**PoD 1**
[完整英文]
> 过渡句

**PoD 2**
[完整英文]
> 过渡句

**PoD 3**
[完整英文]

**DESSERT**
[完整英文]

Relationship with other Skills

	barry-logic-to-speech (this Skill)	barry-logic-training	barry-speech-training	barry-translation-training	quick-english
Positioning	Full pipeline output (orchestrator)	Logical thinking training	Speech structure training	Translation technique training	Urgent quick output
Mode	AI does + user confirms	Coaching / quick	Coaching / quick	Coaching / quick	Direct output
Output	Complete English speech draft	Pyramid structure	Speech framework	English translation	Finished English product

Selection guide：

Wanta complete English speech draft(logic → speech → translation all in one) → barry-logic-to-speech
Only want topractice logic→ barry-logic-training
Only want topractice speech structure→ barry-speech-training
Only want topractice translation→ barry-translation-training
In a hurryWant an English email / impromptu expression → quick-english

Notes

Each stage follows the rules of its corresponding independent skill: the orchestrator does not override the behavioral red lines and coaching rules of independent skills
The fast track does not mean cutting corners: compress confirmation rounds, do not skip the methodology pipeline
Presentation mode: when the user says "I am presenting to students", avoid sensitive information
The pyramid structure must be MECE: groups must be mutually exclusive and collectively exhaustive

About this skill

Positioning: task-driven training entry point (orchestrator)

Training positioning

Related methodology wiki

⚠️ Behavioral red lines (highest priority)

1. Do not repeatedly ask

2. Collect mergeable information all at once

3. AI does the heavy lifting, the user makes the judgments

4. When the user says "just give it to me directly," take the fast track

5. Confirmation gates serve the user, not obstruct the user

Two tracks

Automatic judgment

Standard track

Step 1: Logic organization (call barry-logic-training · organization mode)

Step 2: Speech arrangement (call barry-speech-training · fast mode)

Step 3: Translation output (call barry-translation-training · fast mode)

⭐ Mandatory audience culture cascade constraint (C9 revision)

Step 4: Final delivery

Step 5 (must do after completion): high-frequency chunks into the database + verbatim transcript saved to disk ⭐

Fast track

Step 1: Collect key information all at once

Step 2: AI completes the entire chain in one go

Step 3: Deliver the complete finished product all at once

Step 4 (must-do after completion): high-frequency chunks automatically added to fancy-vocab.md ⭐

A-level special rules (Survival English mode)

Final deliverable template

Relationship with other Skills

Notes