qmd-searchlisted
Install: claude install-skill spikelab/multiplai-cc-mktplace
# qmd-search — manual resources retrieval
Search the user's resources knowledge base through its qmd index. This is
the manual companion to the automatic per-prompt retrieval that
multiplai-context performs when `resources_retrieval=qmd` — use it for
deeper digging, different phrasings, or when the automatic injection
missed a document you expect to exist.
## Resolve the configuration first
The plugin options (in `settings.json` → `pluginConfigs` →
`multiplai-context@<marketplace>` → `options`, or the matching
`CLAUDE_PLUGIN_OPTION_<KEY>` env vars, key uppercased) decide how to reach qmd:
| Option | Default | Meaning |
|---|---|---|
| `qmd_mode` | `local` | `http` = POST to a resident `qmd mcp --http` daemon (preferred); `local` = qmd binary on PATH; `ssh` = qmd runs on the host over the SSH bridge |
| `qmd_http_url` | `http://host.docker.internal:8181` | daemon base URL for `http` mode |
| `qmd_ssh_host` | `host.docker.internal` | bridge host for `ssh` mode |
| `qmd_collection` | `resources` | collection holding the index |
| `resources_dir` | — | maps result URIs back to absolute file paths |
| `workspace_dir` | — | the project-local `.qmd/` index lives at this root |
## Commands (pick by depth)
### `http` mode — POST an authored, typed query (preferred)
Hit the daemon's REST endpoint directly from the container (plain
`curl`, no SSH bridge). **Author the query — do not paste the raw
sentence.** qmd's lexical arm ANDs its terms and carries 2× weight in the
fusion, so