/LEAP

name: LEAP
description: |
  LEAP — 落地执行引擎。内含两条管线：A 分支蒸馏（从 raw data 提取 skill）、
  B 分支融合（多 skill 编织为一个）。被 SkillAlchemy 编排器调用。
  Use when 编排器判断需要蒸馏或融合时。
version: v1.0

LEAP · 落地执行引擎

LEAP 不自己做路由，不跟用户交互。上游 Skill-Alchemy 告诉它走哪条分支，它执行。 所有交互节点由 Skill-Alchemy 编排——LEAP 只负责跑管线，跑完返回结果。

分支路由

| 指令 | 分支 | 管线 | |------|------|------| | `distill` / `蒸馏` | **A 分支** | 蒸馏管线 — 从 raw data 提取 target OS，编译为 persona/tool skill | | `fuse` / `融合` | **B 分支** | 融合管线 — method.skill(骨架) × subject.skill(s)(血肉) → output.skill |

调用模式

| Mode | Trigger | Behavior | |------|---------|----------| | **Full run** | No special keyword | 完整管线 + Gate，输出 skill 包 | | **Plan only** | `只到 Stage 3` 或 `stop_after_stage: 3` | A 分支 Stage 1-3 only，输出 research_plan.json 后停止 | | **Resume** | `从 Stage 4 继续` 或 `resume_from_stage: 4` | A 分支跳过 1-3，使用已有 research_plan.json，跑 4-7 + Gates |

---

A 分支: 蒸馏管线

Source Intake → Intake Assessment → Research Plan Design
  → Research Swarm → Gate 1: Merge
  → Exemplar Discovery → Synthesis (3 agents)
  → Skill Compilation → Gate 2: Validation

Core principle: extract the operating system behind the source, not just the content or answer.

---

A-Stage 1: Source Intake

Input: person, author, method, organization, domain, URL, repo, or local files.

Create package workspace at `output/<target-slug>-skill/`:

output/<target-slug>-skill/
├── README.md
├── SKILL.md.draft
├── references/             # agent reports + exemplars
├── intermediate/           # structured data
├── examples/               # persona: required; tool: optional
└── validation/             # only for deep mode (Phase 8)

> `templates/` 不再预建——输出模板在 LEAP 共享层，产出 skill 不需要。 > `validation/` 只在 depth=deep 且 Phase 8 执行时创建。

Write `intermediate/open_world_task.json` with target, goal, sources, depth_level:

| `depth_level` | Effect | Use case | |:--|:--|:--| | `quick` | Agent count ≤3, skip Phase 8, mark quality: draft | Rapid prototype | | `standard` | No correction to auto-assessment, Phase 8 suggested | Daily use (default) | | `deep` | Agent count upper bound +1 (≤8), Phase 8 required, dual review | Release quality |

---

A-Stage 2: Intake Assessment

Step 1: Source Modality Analysis

Classify every source by what it can reveal:

| Modality | Examples | Reveals | |----------|---------|---------| | `transcript_interview` | podcasts, video captions, Q&A | spontaneous reasoning, analogies, changed positions | | `longform_text` | books, papers, essays, newsletters | core arguments, methodology, narrative structure | | `secondary_criticism` | reviews, biographies, analysis | external perspective, blind spots, competing views | | `video_subtitle` | YouTube, B站 captions | speech patterns, unscripted reasoning | | `social_media` | posts, threads | expression patterns, real-time reactions | | `code_repo` | git repos, PRs | architecture patterns, API contracts, testing strategy |

For each modality present, note what operations it could reveal. Skip absent ones.

Step 2: Domain Inference

Read `domains/<domain>/domain.md` to confirm. Record primary + secondary domains.

Step 3: Evidence Depth Assessment

**Don't count sources — assess their density.**

• ≥3 high-density sources across ≥2 modalities → **rich** (5-8 agents)
• 1-2 high-density sources → **moderate** (3-5 agents)
• 0 high-density, all medium/low → **sparse** (2-3 agents)

Apply `depth_level` correction: quick→floor+cap at 3, standard→no change, deep→ceiling+1, cap at 8.

Step 4: Skill Mode Determination

| Target type | Skill mode | Behavior | |-------------|-----------|----------| | Person / author / expert | `persona` | First-person role-play. Has 角色扮演规则, 身份, 我怎么说话, 决策启发式 | | Domain / method / organization | `tool` | Third-person analytical. Has Activation Rules, Agentic Protocol, Operation Models |

---

A-Stage 3: Research Plan Design

How to select dimensions

1. Read primary domain pack: `domains/<primary-domain>/domain.md` → Research dimensions 2. **If persona, also read `domains/persona-os/domain.md`.** This cross-cutting layer provides OS extraction lenses (decision under constraint, failure processing, value conflict resolution, attention allocation, etc.). 3. Cross dimensions with source modality: match → active, no match → skip 4. Apply evidence depth cap: sparse→merge, moderate→1:1, rich→split 5. Derive new dimensions from source if domain pack menus don't cover what's visible

Self-Check Before Writing research_plan.json

1. **Depth match** — Does agent_count fall within the depth range? 2. **Modality coverage** — Are source_modalities_used actually present? 3. **Dimension coverage** — Is every active dimension covered? 4. **Merge intent** — Deliberate or lazy? Document the rationale.

Search direction must target dilemmas

Good: "What was the hardest decision at [event]? What options did they have?"
Bad:  "What is their leadership style?"

Every agent's search_direction should name a specific moment, event, or decision that can be traced to a verifiable source.

Plan-Only Mode Stop Point

**If invoked with `只到 Stage 3` or `stop_after_stage: 3`:**

写完 `research_plan.json` 后立即停止。输出：

Research plan 已生成，保存在 intermediate/research_plan.json。

[N] 个 agent，维度：
  R1 — [dimension]: [search_direction 摘要]
  R2 — [dimension]: [search_direction 摘要]
  ...

不要进入 Stage 4。等待 Skill-Alchemy 返回确认/调整后的指令。

---

A-Stage 4: Research Swarm

**Resume mode:** 如果在 `从 Stage 4 继续` 模式下调用，直接从已有的 `intermediate/research_plan.json` 读取 agent 配置。跳过 Stage 1-3。

Launch N agents in parallel. Each agent writes `references/R<NN>-<agent_id>.md`:

Status: pass (or warning / fail)

## Key Findings
## Dilemma Decision Cases (≥2 required)
  ### Case N: [one-line summary]
  - 困境 (Dilemma): specific conflict or hard choice
  - 约束 (Constraints): what limited their options
  - 决策步骤 (Decision Steps): what they did, step by step
  - 结果 (Outcome): what happened
  - 可提取的操作 (Extractable Operation): generalizable rule/p