From vicky
Vicky knowledge-base context primer. Loaded automatically on every session start. Tells Claude which MCP tools Vicky exposes, how WORKFLOW.md steers behavior, when to call dashboard/dql, and the active vault layout. Invoke manually with /vicky:setup if context has been compacted away.
How this skill is triggered — by the user, by Claude, or both
Slash command
/vicky:setupThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Loaded automatically at SessionStart. Re-invoke `/vicky:setup` after `/compact` to restore Vicky context.
Loaded automatically at SessionStart. Re-invoke /vicky:setup after /compact to restore Vicky context.
A demand-driven KB stored under vicky/ inside the project root. Conclusions and sources are markdown notes linked with [[wikilinks]]. The Obsidian vault is the project root itself — init scaffolds an .obsidian/ preset (Dataview + graph) at the root and the vicky/ KB subfolder beside it.
/vicky:<name>)| Skill | When |
|---|---|
setup | Auto-fires on SessionStart. Re-invoke after /compact to restore context. |
index [<path>] | Docs → KB sync. Mirrors changed *.md / *.mdx / *.rst / *.txt outside vicky/ into vicky/sources/docs/. Never indexes source code. |
learn | Drain the pending queue into sources and refresh link graphs. No stub conclusions — synthesis is a separate step via conclude. |
research "<topic>" | Topic-focused web research → sources + follow-up questions. Calls learn at the end. |
experiment | Autonomous code-optimisation loop. Reads experiment.md, runs in worktrees, learns time estimates. |
| Tool | Use when |
|---|---|
research-gap "<question>" | Default for any knowledge question. Returns KB context if found, auto-enqueues research when there's a gap. Honors WORKFLOW.md → auto_enqueue. |
query "<question>" | Direct KB lookup, no auto-enqueue. Focus-biased per WORKFLOW.md. |
learn | Drain the pending queue into sources. triage workflow filters to priority: high. |
remember title content | Save findings to vicky/sources/ with frontmatter. Cannot write to conclusions/. |
conclude title content sources | Save a derived conclusion to vicky/conclusions/. sources arg becomes [[wikilinks]] in frontmatter + body. Use once you have a real synthesis. |
enqueue "<question>" | Manually queue a research question. |
relink | Rebuild link graphs + related: frontmatter. |
dashboard | Vault overview (counts, hubs, pending, sources awaiting synthesis, orphans, stale, tags) via Obsidian + Dataview. Call before research sessions and when the user asks about KB state. |
dql "<query>" | Run arbitrary DQL. query="help" returns syntax reference + examples. Use for ad-hoc questions the fixed dashboard doesn't answer. |
web-search | Drive external research when KB has no answer. |
Read vicky/WORKFLOW.md once at session start, again whenever a tool result references a workflow.
Frontmatter keys that affect runtime:
active_focus: [tag, topic] — biases query results, filters dashboard By focus sectionpriority_tags: [perf, blocker] — emphasised in viewsauto_enqueue: true|false — when false, gaps are reported without being queueddefault_workflow: default | deep-dive | triage — triage makes learn only drain priority: high pending notesmin_sources_per_conclusion: N — quality gate hintSections:
Focus — what Vicky cares about right now (bullets)Active Rules — rules every tool call must honourWorkflows — named proceduresRouting — pipe table mapping question regex → workflowEdits to WORKFLOW.md are picked up on the next tool call (mtime cache).
dashboard and dql shell out to obsidian.com vault=<name> eval and require:
OBSIDIAN_CLI env var)If dashboard errors with "Obsidian is not running with the vault "..." open", instruct the user to open the vault in Obsidian first. Do not retry indefinitely — the 10s timeout is intentional.
research-gapdashboarddqlrememberconclude for any source ready to synthesiserelink to refresh the graphDo not call learn reactively for every question — it drains the queue and is expensive. Run it when the user asks, when the queue has accumulated, or per WORKFLOW.md → default_workflow.
The Obsidian vault is the project root. vicky/ is just the KB subfolder. .obsidian/ sits at the vault root, not inside vicky/.
<project root>/ ← Obsidian vault (open this directory in Obsidian)
├── .obsidian/ Dataview + graph + plugin preset (force-written by init)
└── vicky/
├── sources/ external research, papers, web findings
├── conclusions/ synthesised knowledge (only real synthesis — no stubs)
├── pending/ queued research questions
├── .graphify/ semantic graph state (graph.json — read by query_graph)
├── graphs/ Dataview-queryable wiki of the graph (vicky.md + clusters)
├── .graphifyignore graphify scope (excludes pending/, graphs/, etc.)
├── WORKFLOW.md focus + rules + routing (edit to steer)
└── Dashboard.md live Dataview views
Mines projects and conversations into a searchable memory palace. Activates on queries about MemPalace, memory palace, mining, searching, palace setup, wings, rooms, drawers, or recalling past work.
Guides Payload CMS config (payload.config.ts), collections, fields, hooks, access control, APIs. Debugs validation errors, security, relationships, queries, transactions, hook behavior.
Implements vector databases with Pinecone, Weaviate, Qdrant, Milvus, pgvector for semantic search, RAG, recommendations, and similarity systems. Optimizes embeddings, indexing, and hybrid search.
npx claudepluginhub yesitsfebreeze/vicky --plugin vicky