evolve
CommandPropose, evaluate, apply or roll back an evidence-linked skill improvement.
Move a local LLM Wiki to a hosted wiki, verify it, then keep the local folder as a read-only mirror.
> /plugin marketplace add praneybehl/llm-wiki-plugin> /plugin install llm-wiki@llm-wiki
description: Move a local LLM Wiki to a hosted wiki, verify it, then keep the local folder as a read-only mirror. argument-hint: "[<wiki-dir>] [--raw <raw-dir>] [--no-raw]"
Migrate a local wiki to a hosted wiki. Use the `llm-wiki` skill and `skills/llm-wiki/references/hosted-mode.md`.
Arguments (if any): $ARGUMENTS
Set `REMOTE="${CLAUDE_PLUGIN_ROOT}/skills/llm-wiki/scripts/wiki_remote.py"` for the steps below.
1. Resolve `<WIKI_ROOT>` and `<RAW_ROOT>` from the arguments, the nearest project instructions, global agent instructions, or the current project's `wiki/` and `raw/`. Confirm that `<WIKI_ROOT>/SCHEMA.md` exists. Never combine two wikis. 2. Run `python3 "$REMOTE" status`. If no connection exists, run the token branch of `/wiki:init --hosted <url>` first. The import uses `wiki_remote.py`, so it needs a token also when my MCP client signs in with OAuth. The token needs `write` scope, and `raw:write` for raw files. An empty wiki is the normal target. If the remote wiki already has pages, tell me and ask before you continue. 3. Run local lint once: `python "${CLAUDE_PLUGIN_ROOT}/skills/llm-wiki/scripts/wiki_lint.py" "<WIKI_ROOT>"`. Keep the finding counts for step 7. 4. Dry run: `python3 "$REMOTE" import "<WIKI_ROOT>" --raw "<RAW_ROOT>" --dry-run`. Use `--no-raw` if I asked for it or if the server has no raw storage. Show me the counts: pages to upload, pages that are identical, raw files, bytes, and any refusals. 5. Ask for approval. Then run the same command without `--dry-run`. The import uploads in small batches. It skips identical files. If it stops, run the same command again. It continues from where it stopped. 6. Read the JSON report.
7. Verify on the server: run `python3 "$REMOTE" lint` and compare the counts with step 3. Run two or three searches that I know the answers to with `python3 "$REMOTE" search "<query>"`. Report the differences. Small search ranking differences are normal. 8. Ask whether to keep the local folder as a read-only mirror. If yes, explain that the next pull replaces local files with the server copy and saves a changed local file as `<file>.local-<timestamp>`. After my approval, run `python3 "$REMOTE" pull "<WIKI_ROOT>"`. Add `--git` if `<WIKI_ROOT>` is in a git repository and I want a commit for each pull. 9. Propose the hosted-wiki stanza from `references/agent-memory-integration.md` to replace the local stanza. Show the exact change. Write it only after I approve. 10. Tell me that the local folder is now a mirror: edits there do not go to the server. `/wiki:learn` and `/wiki:evolve` can run on the mirror. `python3 "$REMOTE" export <file>.tar.gz` makes a full backup at any time.
Rollback: until step 8, the local wiki and agent instructions do not change. To stop, keep using the local wiki.
Turn PDFs, articles, transcripts, and notes into a shared wiki that your AI agents can search, cite, and keep up to date. Add a source once. Ask questions later. Keep the useful answers.
Propose, evaluate, apply or roll back an evidence-linked skill improvement.
Ingest a source (paper, article, transcript, PDF, notes) into the LLM Wiki.
Initialize a personal global LLM Wiki or one in this project, or connect this agent to a hosted wiki.
Capture verified task experience and consolidate reusable lessons in the wiki.
Compare two or more options side by side, recommend one, and let the reader pick in the page
Explain code in detail - what it does, how it works, and why
Create smart commit with quality checks and conventional format