Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

feat(agent-plugins): add ov-resources skill (#6092)

Zayn Jarvis6006fc3403b5
diff --git a/agent-plugins/README.md b/agent-plugins/README.md
index 93e11a7e8..89004e5f0 100644
--- a/agent-plugins/README.md
+++ b/agent-plugins/README.md
@@ -16,6 +16,7 @@ skills/ov-memory-troubleshoot/ # read-only extraction troubleshooting
skills/ov-experience-memory/ # retrieve and apply prior task Experience
skills/openviking-skills/ # find, use, create, and share OpenViking skills
skills/ov-kanban/ # kanban board: pick, execute, hand off tasks stored as OV markdown files
+skills/ov-resources/ # search and read documents stored under viking://resources/
plugin.test.mjs # node --test conformance checks
```
@@ -34,7 +35,7 @@ Zero npm dependencies; the proxy and tests run on the Node.js standard library (
1. Have an OpenViking server reachable (see the [quickstart](../docs/en/getting-started/02-quickstart.md)); default local endpoint is `http://127.0.0.1:1933`.
2. Point your Agent-Plugins-conforming client at this directory (each client has its own install command or plugin directory; consult its docs). The client will:
- register the `openviking` MCP server from `mcp.json` — it runs `node <plugin>/servers/mcp-proxy.mjs` over stdio;
- - discover the `openviking-memory`, `ov-memory-troubleshoot`, `ov-experience-memory`, `openviking-skills`, and `ov-kanban` skills from `skills/`.
+ - discover the `openviking-memory`, `ov-memory-troubleshoot`, `ov-experience-memory`, `openviking-skills`, `ov-kanban`, and `ov-resources` skills from `skills/`.
3. Configure credentials (next section) and start a session. The model gains `find` / `search` / `read` / `remember` / `write` / `add_skill` and the other OpenViking MCP tools. Use `search` with `mode="context"` for server-assembled context.
## Why a stdio proxy instead of a `streamable-http` entry
diff --git a/agent-plugins/skills/ov-resources/SKILL.md b/agent-plugins/skills/ov-resources/SKILL.md
new file mode 100644
index 000000000..d5694675a
--- /dev/null
+++ b/agent-plugins/skills/ov-resources/SKILL.md
@@ -0,0 +1,53 @@
+---
+name: ov-resources
+description: Search and read documents, repos, URLs, and other resources stored in OpenViking under viking://resources/. Use whenever the user asks about docs, specs, codebases, papers, or notes that were added to OpenViking ("check the docs", "what does our design doc say", "find it in OpenViking"), or when a task needs reference material this session does not have. Covers picking between find, grep, glob, and list, and drilling from a hit down to the exact file.
+---
+
+# OpenViking Resources
+
+Resources are documents a user added to OpenViking (files, repos, URLs).
+They live in a file tree under `viking://resources/` (and sometimes
+`viking://user/<id>/resources/` for user-private ones). Each directory has two
+summaries next to its files:
+
+- `.abstract.md` — one paragraph: what the directory holds (L0)
+- `.overview.md` — a structured overview of its contents (L1)
+
+Read summaries first and open full files last. The summaries exist so you can
+decide where to look without reading everything.
+
+## Pick the tool
+
+| You know… | Use |
+|---|---|
+| roughly what it is about | `find(query, target_uri="viking://resources")` |
+| an exact word, name, or error string | `grep(uri, pattern=[...])` |
+| a file name or extension | `glob(pattern, uri)` e.g. `**/*.pdf` |
+| where it is | `list(uri)` and `read(uris)` |
+
+Default to `find`. Scope it with `target_uri` as soon as you know which
+resource you are in — narrower scopes give better hits. Use `search` instead
+of `find` only when the question is vague and needs intent analysis.
+
+## Drill down
+
+1. `find` returns URIs with a score. A hit on `<dir>/.abstract.md` means the
+ whole directory matched, not that the answer is in the abstract.
+2. `read` that directory's `.overview.md` to see what is inside.
+3. `list` the directory, or run a scoped `find`/`grep` inside it, to reach the
+ specific file.
+4. `read` the file. For long files, `grep` first and then `read` with
+ `offset`/`limit` around the match.
+
+Stop as soon as you have what the task needs. Two or three targeted reads
+beat ten broad ones.
+
+## Tips
+
+- If `find` returns nothing useful, `list viking://resources` to see what
+ exists, then search inside the closest directory.
+- Low scores (under ~0.4) are usually noise; rephrase with the domain's
+ own terms rather than raising `limit`.
+- Cite the `viking://` URI of what you used so the user can open it.
+- To add a new resource, use `add_resource` with the file path or URL; do not
+ copy its text in with `write`.