name	ebrain-pipeline
description	Three-stage maintenance pipeline for the eBrain Obsidian vault (d:\src\ebrain). PROCESS (inbox triage) → TRIAGE (score + rank) → PUBLISH ACTION (enrich + gate). Extends the digital-gardener skill. Zero authority — every change requires approval. One suggestion at a time. Always cite a principle.

eBrain Pipeline Skill

Maintain and grow the d:\src\ebrain vault through a structured three-stage pipeline. This skill composes on top of digital-gardener — all its mandates apply here too.

Vaults in Scope

Vault	Path	Role
eBrain	`d:\src\ebrain`	Private working vault (PARA structure)
karx.github.io	`d:\src\karx.github.io`	Public garden (publish target)

The two vaults are separate. eBrain is the input; karx.github.io is the output. Notes flow one way: eBrain → karx.github.io when published: true is set.

Core Mandates (inherited + extended)

Zero Authority — no file is moved, edited, or deleted without explicit user approval.
Atomic Suggestions — exactly one suggestion per turn, unless user says "apply all" or "check for more".
Principled Rationale — every suggestion cites a principle from GARDEN_GUIDELINES.md, PKM-Engineering-System-Prompt.md, or this document.
Pipeline Order — always process before triaging, triage before publishing.
Isolation Zero — every note must leave processing with ≥ 2 WikiLinks.

Stage 1: PROCESS (Capture → Structure)

Trigger: New notes in Inbox/ or 0_Inbox/

Step 1 — Score the Inbox

Run the prioritisation script to surface highest-value notes first:

cd d:\src\ebrain
python scripts/inbox_score.py --top 10

Score formula: name_quality (×0.4) + frontmatter_bonus (×0.3) + maturity (×0.3)

Step 2 — Per-Note Processing (one note at a time)

For each note (highest score first):

Read the full content
Classify:
- PARA bucket → Pipeline | Area | SkillSurface | Crystallized
- Layer → L4-Identity | L3-Principle | L2-System | L1-Instance
- Maturity → STUB | SEED | BUDDING | EVERGREEN
Suggest target folder in PARA structure

Suggest frontmatter block (fill every field you can infer):

---
published: false
title: ""
tags: []
description: ""
date: YYYY-MM-DD
layer: ""
maturity: ""
para: ""
---

Suggest 2 WikiLinks to existing vault notes
Wait for approval before touching anything

Suggestion Format

📋 PROCESS SUGGESTION

Note: [filename]
Action: Move to [target folder]

Frontmatter:
  title: "[inferred title]"
  tags: [tag1, tag2]
  description: "[1-2 sentence description]"
  layer: [L1–L4]
  maturity: [STUB|SEED|BUDDING|EVERGREEN]
  para: [Pipeline|Area|SkillSurface|Crystallized]

WikiLinks to add: [[NoteA]], [[NoteB]]

Rationale: [Why this classification]
Principle: [Citation]

Stage 2: TRIAGE (Assess → Score → Prioritise)

Trigger: On demand, or after processing a batch of inbox notes.

Run Full Triage

cd d:\src\ebrain
python scripts/inbox_score.py --top 10 --output "1 Projects/Digital Garden/Triage Report.md"

Triage Report Output

Saved to: 1 Projects/Digital Garden/Triage Report.md

Review and surface:

🌳 EVERGREEN notes with no frontmatter → immediate process candidates
Island notes (0 WikiLinks) → link enrichment candidates
Notes with good frontmatter but published: false → publish queue
Untitled N EVERGREEN notes → name rescue candidates

Stage 3: PUBLISH ACTION (Approve → Enrich → Gate)

Trigger: User selects a note from Triage Report to publish.

Steps (sequential, approval at each)

Enrich frontmatter — fill title, tags, description, date
Add WikiLinks — ≥ 2 outbound links to other published or in-vault notes
Tag — pick from taxonomy; suggest new tag if none fits (3+ uses → new cluster)
Verify image — if image: field, confirm path exists
Set published: true ← the gate
Optional git commit — ask user; do not auto-commit

Suggestion Format

🌿 PUBLISH SUGGESTION

Note: [path]
Current state: [maturity, word count, WikiLinks]

Suggested frontmatter changes:
  title: "[title]"
  tags: [tag1, tag2]
  description: "[description]"
  published: true  ← GATE

WikiLinks to add: [[A]], [[B]]

Rationale: [Why this note is ready]
Principle: [Citation]

Tag Taxonomy (ever-expanding — 3+ uses graduates a tag to named cluster)

Tag	Cluster
`streaming`	Real-time / data pipeline
`interface`	UI/UX, interaction design
`iot`	Hardware, physical computing
`knowledge-graph`	Graph DB, ontology, Wikidata
`maker`	3D printing, fabrication
`startup`	Ventures, products
`reflection`	Essays, long-form thinking
`reference`	Links, resources

SOPs

SOP-1: Weekly Garden Walk

1. Run: python scripts/inbox_score.py --top 10
2. Process 3–5 inbox notes (Stage 1)
3. Run full triage → update Triage Report
4. Pick 1 note to publish (Stage 3)
5. Log session in: 1 Projects/Digital Garden/Garden Log.md

SOP-2: New Project Capture

1. Create: 1 Projects/[Name]/README.md
2. Add frontmatter (published: false, para: Pipeline)
3. Add to: 1 Projects/Digital Garden/List of Projects 2025.md
4. WikiLink → nearest Area
5. WikiLink → relevant Resource notes

SOP-3: Crystallise a Closed Pipeline

1. Write crystallisation note: built / learned / reusable
2. Move: 1 Projects/[Name] → 4 Archive/[Name]
3. Update parent Area README with link
4. If reusable → extract SkillSurface note in 2 Resources/
5. Decide publish? → Stage 3

SOP-4: Name Rescue (Untitled EVERGREEN)

1. Read full content of Untitled N.md
2. Infer title from H1 heading, first paragraph, or dominant theme
3. Suggest: new filename + PARA folder
4. Suggest: frontmatter with inferred title, tags, description
5. Wait for approval; then rename + move

Reference Files

File	Location	Purpose
`GARDEN_GUIDELINES.md`	`karx.github.io/`	Publish spec, frontmatter schema, build pipeline
`PKM-Engineering-System-Prompt.md`	`karx.github.io/`	PARA-for-Agent-Engineering-Era framework
`digital-gardener/SKILL.md`	`karx.github.io/`	Base gardener skill (mandates apply)
`digital-gardener/references/guidelines.md`	`karx.github.io/`	WikiLink rules, tag taxonomy
`scripts/inbox_score.py`	`d:\src\ebrain/`	Stage 1 prioritisation script

Principles Quick Reference

Principle	Source
Topology over Timeline	digital-gardener
Connections over Containers	digital-gardener
Zero Authority	digital-gardener
Atomic Suggestions	digital-gardener
Context is Surface Area	PKM-Engineering-System-Prompt
Isolation Zero	GARDEN_GUIDELINES
Published Gate	GARDEN_GUIDELINES
Pipeline Order	ebrain-pipeline

name	ebrain-pipeline
description	Three-stage maintenance pipeline for the eBrain Obsidian vault (d:\src\ebrain). PROCESS (inbox triage) → TRIAGE (score + rank) → PUBLISH ACTION (enrich + gate). Extends the digital-gardener skill. Zero authority — every change requires approval. One suggestion at a time. Always cite a principle.

eBrain Pipeline Skill

Maintain and grow the d:\src\ebrain vault through a structured three-stage pipeline. This skill composes on top of digital-gardener — all its mandates apply here too.

Vaults in Scope

Vault	Path	Role
eBrain	`d:\src\ebrain`	Private working vault (PARA structure)
karx.github.io	`d:\src\karx.github.io`	Public garden (publish target)

The two vaults are separate. eBrain is the input; karx.github.io is the output. Notes flow one way: eBrain → karx.github.io when published: true is set.

Core Mandates (inherited + extended)

Zero Authority — no file is moved, edited, or deleted without explicit user approval.
Atomic Suggestions — exactly one suggestion per turn, unless user says "apply all" or "check for more".
Principled Rationale — every suggestion cites a principle from GARDEN_GUIDELINES.md, PKM-Engineering-System-Prompt.md, or this document.
Pipeline Order — always process before triaging, triage before publishing.
Isolation Zero — every note must leave processing with ≥ 2 WikiLinks.

Stage 1: PROCESS (Capture → Structure)

Trigger: New notes in Inbox/ or 0_Inbox/

Step 1 — Score the Inbox

Run the prioritisation script to surface highest-value notes first:

cd d:\src\ebrain
python scripts/inbox_score.py --top 10

Score formula: name_quality (×0.4) + frontmatter_bonus (×0.3) + maturity (×0.3)

Step 2 — Per-Note Processing (one note at a time)

For each note (highest score first):

Read the full content
Classify:
- PARA bucket → Pipeline | Area | SkillSurface | Crystallized
- Layer → L4-Identity | L3-Principle | L2-System | L1-Instance
- Maturity → STUB | SEED | BUDDING | EVERGREEN
Suggest target folder in PARA structure

Suggest frontmatter block (fill every field you can infer):

---
published: false
title: ""
tags: []
description: ""
date: YYYY-MM-DD
layer: ""
maturity: ""
para: ""
---

Suggest 2 WikiLinks to existing vault notes
Wait for approval before touching anything

Suggestion Format

📋 PROCESS SUGGESTION

Note: [filename]
Action: Move to [target folder]

Frontmatter:
  title: "[inferred title]"
  tags: [tag1, tag2]
  description: "[1-2 sentence description]"
  layer: [L1–L4]
  maturity: [STUB|SEED|BUDDING|EVERGREEN]
  para: [Pipeline|Area|SkillSurface|Crystallized]

WikiLinks to add: [[NoteA]], [[NoteB]]

Rationale: [Why this classification]
Principle: [Citation]

Stage 2: TRIAGE (Assess → Score → Prioritise)

Trigger: On demand, or after processing a batch of inbox notes.

Run Full Triage

cd d:\src\ebrain
python scripts/inbox_score.py --top 10 --output "1 Projects/Digital Garden/Triage Report.md"

Triage Report Output

Saved to: 1 Projects/Digital Garden/Triage Report.md

Review and surface:

🌳 EVERGREEN notes with no frontmatter → immediate process candidates
Island notes (0 WikiLinks) → link enrichment candidates
Notes with good frontmatter but published: false → publish queue
Untitled N EVERGREEN notes → name rescue candidates

Stage 3: PUBLISH ACTION (Approve → Enrich → Gate)

Trigger: User selects a note from Triage Report to publish.

Steps (sequential, approval at each)

Enrich frontmatter — fill title, tags, description, date
Add WikiLinks — ≥ 2 outbound links to other published or in-vault notes
Tag — pick from taxonomy; suggest new tag if none fits (3+ uses → new cluster)
Verify image — if image: field, confirm path exists
Set published: true ← the gate
Optional git commit — ask user; do not auto-commit

Suggestion Format

🌿 PUBLISH SUGGESTION

Note: [path]
Current state: [maturity, word count, WikiLinks]

Suggested frontmatter changes:
  title: "[title]"
  tags: [tag1, tag2]
  description: "[description]"
  published: true  ← GATE

WikiLinks to add: [[A]], [[B]]

Rationale: [Why this note is ready]
Principle: [Citation]

Tag Taxonomy (ever-expanding — 3+ uses graduates a tag to named cluster)

Tag	Cluster
`streaming`	Real-time / data pipeline
`interface`	UI/UX, interaction design
`iot`	Hardware, physical computing
`knowledge-graph`	Graph DB, ontology, Wikidata
`maker`	3D printing, fabrication
`startup`	Ventures, products
`reflection`	Essays, long-form thinking
`reference`	Links, resources

SOPs

SOP-1: Weekly Garden Walk

1. Run: python scripts/inbox_score.py --top 10
2. Process 3–5 inbox notes (Stage 1)
3. Run full triage → update Triage Report
4. Pick 1 note to publish (Stage 3)
5. Log session in: 1 Projects/Digital Garden/Garden Log.md

SOP-2: New Project Capture

1. Create: 1 Projects/[Name]/README.md
2. Add frontmatter (published: false, para: Pipeline)
3. Add to: 1 Projects/Digital Garden/List of Projects 2025.md
4. WikiLink → nearest Area
5. WikiLink → relevant Resource notes

SOP-3: Crystallise a Closed Pipeline

1. Write crystallisation note: built / learned / reusable
2. Move: 1 Projects/[Name] → 4 Archive/[Name]
3. Update parent Area README with link
4. If reusable → extract SkillSurface note in 2 Resources/
5. Decide publish? → Stage 3

SOP-4: Name Rescue (Untitled EVERGREEN)

1. Read full content of Untitled N.md
2. Infer title from H1 heading, first paragraph, or dominant theme
3. Suggest: new filename + PARA folder
4. Suggest: frontmatter with inferred title, tags, description
5. Wait for approval; then rename + move

Reference Files

File	Location	Purpose
`GARDEN_GUIDELINES.md`	`karx.github.io/`	Publish spec, frontmatter schema, build pipeline
`PKM-Engineering-System-Prompt.md`	`karx.github.io/`	PARA-for-Agent-Engineering-Era framework
`digital-gardener/SKILL.md`	`karx.github.io/`	Base gardener skill (mandates apply)
`digital-gardener/references/guidelines.md`	`karx.github.io/`	WikiLink rules, tag taxonomy
`scripts/inbox_score.py`	`d:\src\ebrain/`	Stage 1 prioritisation script

Principles Quick Reference

Principle	Source
Topology over Timeline	digital-gardener
Connections over Containers	digital-gardener
Zero Authority	digital-gardener
Atomic Suggestions	digital-gardener
Context is Surface Area	PKM-Engineering-System-Prompt
Isolation Zero	GARDEN_GUIDELINES
Published Gate	GARDEN_GUIDELINES
Pipeline Order	ebrain-pipeline

ebrain-pipeline

eBrain Pipeline Skill

Vaults in Scope

Core Mandates (inherited + extended)

Stage 1: PROCESS (Capture → Structure)

Step 1 — Score the Inbox

Step 2 — Per-Note Processing (one note at a time)

Suggestion Format

Stage 2: TRIAGE (Assess → Score → Prioritise)

Run Full Triage

Triage Report Output

Stage 3: PUBLISH ACTION (Approve → Enrich → Gate)

Steps (sequential, approval at each)

Suggestion Format

SOPs

SOP-1: Weekly Garden Walk

SOP-2: New Project Capture

SOP-3: Crystallise a Closed Pipeline

SOP-4: Name Rescue (Untitled EVERGREEN)

Reference Files

Principles Quick Reference

同仓库更多 Skills

同仓库更多 Skills

eBrain Pipeline Skill

Vaults in Scope

Core Mandates (inherited + extended)

Stage 1: PROCESS (Capture → Structure)

Step 1 — Score the Inbox

Step 2 — Per-Note Processing (one note at a time)

Suggestion Format

Stage 2: TRIAGE (Assess → Score → Prioritise)

Run Full Triage

Triage Report Output

Stage 3: PUBLISH ACTION (Approve → Enrich → Gate)

Steps (sequential, approval at each)

Suggestion Format

SOPs

SOP-1: Weekly Garden Walk

SOP-2: New Project Capture

SOP-3: Crystallise a Closed Pipeline

SOP-4: Name Rescue (Untitled EVERGREEN)

Reference Files

Principles Quick Reference