triage-ideas
GitHub处理Obsidian Ideas.md收件箱，将想法分类并路由到对应项目路线图。支持识别类型、优先级和工时，经用户确认后更新本地文件及数据库，避免直接编辑路线图。
Trigger Scenarios
Install
npx skills add NeverSight/learn-skills.dev --skill triage-ideas -g -y
SKILL.md
Frontmatter
{
"name": "triage-ideas",
"description": "Process Obsidian Ideas.md inbox — route ideas to project roadmaps. NOT for direct roadmap edits or reviews.",
"user-invocable": true
}
Triage Ideas — Obsidian Inbox to Project Roadmaps
Process the Ideas inbox — route each idea to the appropriate project roadmap.
Obsidian CLI (for read/search only — no edit command):
OBS="<path-to-obsidian-binary>"
# All commands need 2>&1 | grep -v "Loading\|out of date"
Obsidian vault path: {{vault_path}}
Roadmap file: {{vault_path}}/{{project}}/Roadmap.md
Ideas file: {{vault_path}}/Ideas.md
Step 1 — Read the inbox
"$OBS" read path="Ideas.md" 2>&1 | grep -v "Loading\|out of date"
If the only content is the frontmatter and # Ideas header with no bullets, report "Ideas inbox is clear" and stop.
Step 2 — Classify each idea
Analyze each idea and classify:
- Project: Which project does this belong to?
- Type: Roadmap item (future work), Todo (actionable now), or Feature request
- Priority: BLOCKING, CRITICAL, High, Medium, Low, or "When needed"
- Effort: Low, Medium, Large
- Table: Immediate Backlog (actionable items) or Long-Term Ideas (exploratory/future)
Use vault context to flesh out the idea (run searches in parallel):
"$OBS" search query="{relevant keywords}" limit=10 2>&1 | grep -v "Loading\|out of date"
Also read the roadmap to check for duplicates.
Present the triage plan to the user before executing:
| # | Idea | Project | Type | Priority | Effort | Table |
|---|------|---------|------|----------|--------|-------|
| 1 | ... | Project A | Roadmap | Medium | Low | Immediate |
| 2 | ... | Project A | Roadmap | Low | Medium | Long-Term |
Ask the user to confirm or adjust before proceeding.
Step 3 — Route to ALL destinations
For each confirmed idea, update destinations:
3a. Obsidian Roadmap.md
Use the Read tool then Edit tool on the actual file path (NOT Obsidian CLI — it has no edit command):
- Immediate Backlog items: Insert new row in the Immediate table
- Long-Term Ideas items: Append to the Long-Term Ideas table
Row format:
| {Priority} | **{Title}** — {Description} | Ideas inbox {YYYY-MM-DD} | {Effort} |
3b. Database (if applicable)
If using a database for roadmap tracking (e.g., D1, SQLite), batch ALL ideas into a single SQL file to minimize round-trips:
cat > tmp-ideas.sql << 'EOSQL'
INSERT INTO roadmap_items (id, title, description, project, status, createdDate, createdBy, updatedAt, updatedBy, estimatedTime)
VALUES
('idea-xxx', 'Title 1', 'Description. Priority: Medium.', 'your-project', 'planned', 'YYYY-MM-DD', 'claude', 'YYYY-MM-DDTHH:MM:SSZ', 'claude', 'Medium'),
('idea-yyy', 'Title 2', 'Description. Priority: Low.', 'your-project', 'planned', 'YYYY-MM-DD', 'claude', 'YYYY-MM-DDTHH:MM:SSZ', 'claude', 'Large');
EOSQL
# Execute against your database(s)
# npx wrangler d1 execute your-db --remote --file=tmp-ideas.sql
rm tmp-ideas.sql
3c. Other project roadmaps
For items routed to other projects, use the Edit tool on:
{{vault_path}}/{Project}/Roadmap.md
If no Roadmap.md exists, create one with the Write tool:
---
type: roadmap
last_updated: {today}
---
# {Project} — Roadmap
## Backlog
| Priority | Item | Notes | Effort |
|----------|------|-------|--------|
## Completed
Step 4 — Clear the inbox
Use the Read tool then Edit tool on {{vault_path}}/Ideas.md.
Replace all idea content, leaving only:
---
type: ideas
---
# Ideas
IMPORTANT: Do NOT use Obsidian CLI for editing — it doesn't support it. Always use Read + Edit tools on the actual file path.
Step 5 — Report summary
## Ideas Triaged
Processed X ideas from inbox:
- → ProjectA/Roadmap.md (Immediate): "idea 1", "idea 2"
- → ProjectA/Roadmap.md (Long-Term): "idea 3"
- → ProjectB/Roadmap.md: "idea 4"
- Marked as already done: "idea 5"
Ideas inbox cleared.
Gotchas
- Obsidian CLI has NO edit command — always use Read + Edit tools on full file paths
- Batch database inserts — combine all ideas into one SQL file, run against all environments
- Database has no priority column — encode priority in the description text if needed
- Dedup check — search the roadmap for similar items before adding duplicates
- Immediate vs Long-Term — actionable items with clear scope go to Immediate Backlog; exploratory/meta/tooling ideas go to Long-Term Ideas
- Clean up tmp files — always remove temp SQL files after execution
Version History
- e0220ca Current 2026-07-05 23:34
