Getnote Search
Pull semantically ranked excerpts from Get笔记 notes and knowledge bases through the getnote CLI while an agent is researching or coding.
Overview
getnote-search is an agent skill most often used in Idea (also Build, Grow) that runs semantic search over Get笔记 notes via the getnote CLI and returns ranked excerpts for agents.
Install
npx skills add https://github.com/iswalle/getnote-cli --skill getnote-searchWhat is this skill?
- Runs `getnote search <query>` with optional `--kb <topic_id>` and `--limit` (default 10, maximum 10)
- Returns results ranked by semantic relevance with note_id, title, content excerpt, score, created_at, and note_type
- Supports `-o json` with top-level `results` array for programmatic parsing by agents
- Documents that `note_id` is populated only for NOTE-type hits; FILE, BLOGGER, LIVE, and other types return empty note_id
- Pairs with `getnote kbs -o json` to resolve `topic_id` for scoped knowledge-base search
- Default and maximum `--limit` is 10 results per search
- Results include semantic relevance `score` ranked high to low
- `note_id` is only set for NOTE-type results; other note types return empty note_id
Adoption & trust: 627 installs on skills.sh; 127 GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
Your agent cannot see the notes and knowledge bases you already captured in Get笔记, so answers drift away from your own research.
Who is it for?
Builders who already use Get笔记, have authenticated the getnote CLI, and want programmatic semantic recall during research, implementation, or content work.
Skip if: Teams without Get笔记 or getnote installed, or workflows that need full note bodies without a separate fetch when only excerpts are returned.
When should I use this skill?
You need semantically ranked excerpts from Get笔记 (all notes or one knowledge base) via authenticated `getnote search`, especially when parsing `-o json` in an agent workflow.
What do I get? / Deliverables
You get up to ten semantically ranked note excerpts—optionally scoped to one knowledge base—with JSON-friendly fields so the agent can follow up on NOTE rows or adjust when note_id is empty.
- Ranked search result list (CLI or JSON with top-level `results`)
- Per-hit metadata: title, content excerpt, score, created_at, note_type, and note_id when type is NOTE
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Canonical shelf is Idea → research because solo builders most often search a personal knowledge base while exploring topics, competitors, and prior captures before committing to build. The research subphase is where note retrieval replaces ad-hoc memory—semantic search over stored notes supports discovery and grounding early in the journey.
Where it fits
Run `getnote search "大模型 API" -o json` to pull prior API comparisons before choosing a vendor.
Search within one KB with `--kb` to see what you already documented about MVP features before rewriting a spec.
Query `getnote search "RAG" --limit 5` for implementation notes while wiring retrieval in your app.
Semantic-search marketing drafts stored as notes to reuse angles for a launch post.
How it compares
Use as a Get笔记 CLI integration for personal knowledge retrieval—not as a general web search skill or an MCP server wrapper.
Common Questions / FAQ
Who is getnote-search for?
Solo and indie builders who store research and references in Get笔记 and want Claude Code, Cursor, Codex, or similar agents to query those notes through the getnote CLI.
When should I use getnote-search?
During Idea research to surface prior captures, while validating scope from existing notes, during Build when looking up technical references in a KB, and in Grow when recycling content ideas—especially with `-o json` for agent parsing.
Is getnote-search safe to install?
Treat it like any third-party CLI skill: review the Security Audits panel on this Prism page and confirm getnote auth and network behavior match your policy before running searches on production machines.
SKILL.md
READMESKILL.md - Getnote Search
# getnote-search Skill Semantic search across all notes or within a specific knowledge base. ## Prerequisites - `getnote` CLI installed and authenticated (`getnote auth status` should show "Authenticated") ## Commands ### Search notes ``` getnote search <query> [--kb <topic_id>] [--limit <n>] ``` | Flag | Default | Description | |------|---------|-------------| | `--kb` | — | Limit search to a knowledge base (`topic_id`) | | `--limit` | 10 | Max results (max 10) | Results are ranked by semantic relevance (high → low). Each result includes: `note_id`, `title`, `content` (excerpt), `score`, `created_at`, `note_type`. > Note: `note_id` is only populated for `NOTE` type results. Other types (`FILE`, `BLOGGER`, `LIVE`, etc.) return an empty `note_id`. ```bash # Search across all notes getnote search "大模型 API" # Search within a knowledge base getnote search "RAG" --kb qnNX75j0 # Limit results + JSON output getnote search "机器学习" --limit 5 -o json ``` --- ## Agent Usage Notes - Use `-o json` when parsing results programmatically. - JSON response: `{"success":true,"results":[{"note_id":"...","title":"...","content":"...","score":0.95,"created_at":"...","note_type":"..."}]}` - Note: `results` is at the top level, not nested under `data`. - Get `topic_id` for `--kb` from `getnote kbs -o json` → `data.topics[].topic_id`. - For `NOTE` type results, use `getnote note <note_id>` to get the full content. - Max `--limit` is 10; use `getnote notes` for browsing without a query. - Exit code `0` = success; non-zero = error. Error details go to stderr.