Stop losing decisions, action items, and brilliant ideas to the void. core-skills turns every conversation into structured, searchable knowledge -- and distills it into a crosslinked knowledge wiki your AI answers from. Use slash commands in Claude Code, or describe what you need in Cowork.
The latest additions to the core-skills ecosystem
A canonical people roster plus a folder-memory check: every name in a new summary is compared against what the folder has already published, so one person can no longer end up spelled three different ways across a meeting series -- and a plausible mis-hearing can't quietly replace an established term.
Every meeting type now has a shape contract -- heading order, table columns, decision marker -- checked before each save. /ops lint shows exactly where an existing series forked. Format changes become explicit contract edits instead of accidental drift.
Document systems rot quietly: indexes lag, task ledgers freeze, sent material never gets archived. /ops sweep finds all of it in one read-only pass and offers the fixes -- and anything that moves now leaves a pointer at its old home. New in v1.27: the ecosystem's own version-alignment check joined the sweep, so even the tooling can no longer drift unnoticed. v1.29 adds structure conformance: stray or variant inbox/outbox folders anywhere in the tree are found before they can fork the pending list.
One human-owned markdown file where quick notes land and get sorted into today / this week / later. Meeting preps automatically pull your open items for that contact, the dashboard shows today's priorities, and /inbox triage refresh handles the mechanical upkeep -- without ever reordering your thinking. And since v1.28 the inbox door also takes file drops: a PDF or CSV dropped in _inbox/.files/ is processed and travels with its output to its real home.
Same skills, same results. Choose the interface that fits you.
For developers and power users. Slash commands in the terminal -- type /ops, paste a transcript, done.
For everyone else. Describe what you need in plain language in Claude Desktop -- no terminal required.
How to use Cowork
Use slash commands in Claude Code or describe tasks in Cowork. No browser tabs, no SaaS dashboards. Just Claude and your vault.
Claude reads your transcript, identifies participants, extracts decisions, creates summaries, and offers to import action items -- all in one go. You review and approve. That's it.
Over weeks and months, insights accumulate. Decisions get tracked. Patterns emerge. Your vault becomes a searchable institutional memory that gets smarter with every conversation.
What changes when every meeting has a memory
Each skill handles one part of the workflow. They share standards, connect data, and work together without configuration.
The main meeting processor. Paste a transcript and get structured summaries, CHANGELOG entries, task imports, and knowledge extraction -- all controlled by organization config. Run /ops status to see available configs.
Universal content capture. Voice memos, quick notes, links, ideas -- anything goes. /inbox classifies, saves to _inbox/, and suggests where to route it next.
For personal calls and ad-hoc recordings when you don't need org config. Clean summary, CHANGELOG, and knowledge extraction.
Auto-generates meeting briefings from previous conversations. Open items, talking points, a suggested agenda, and cross-references from other contacts -- ready in seconds.
Distributed task tracker with per-folder files, priorities, source linking, and automatic carry-forward. Tasks from meetings show up in your dashboard.
Start your morning with today's meetings, active tasks, and quick-access symlinks. Auto-discovers preparations and completed calls.
Knowledge extraction and skill evolution. Backfills insights, compiles execution feedback into patterns, proposes skill improvements. Subcommands: reprocess, compile, propose.
Manages skill repos, symlinks, and updates with multi-remote version safety. One command keeps everything current.
From your first coffee to wrapping up -- here's how the skills fit into a real workday.
See today's meetings, active tasks, and preparations. Desktop symlinks give you one-click access to everything you need.
Claude reads your previous conversations with David, pulls open items, and creates a briefing. You walk in knowing exactly where you left off.
Paste the transcript. Get a structured summary, decisions tracked, action items imported, and insights extracted. The meeting lives on in your vault.
Check your tasks across all projects. Mark items done. Everything syncs to tomorrow's dashboard automatically.
Every conversation contains hidden value. core-skills extracts it silently, builds a knowledge base over time -- and synthesizes it into a wiki of crosslinked topic articles with a master index your AI reads first. Ask “what have I learned about X?” and get a sourced answer. No vector database, no retrieval pipeline.
"Chose Flutter over native for shared codebase"
"Prefers async updates over daily standups"
"Two-week sprints too long -- switch to weekly"
"Could build SaaS from internal tool"
"Budget deferred three meetings in a row"
The companion web app reads your vault in real-time and turns it into an interactive dashboard with AI-powered synthesis.
Day bands, insight cards, synthesis summaries
Triage, classify, route content
Kanban view with priority filters
Ops files, meetings, contacts per project
Browse vault files with preview
Type x tag pivot heatmap with drill-down
Activity heatmap and trend charts
Claude queries vault via JSON-RPC
Cross-project synthesis works with your choice of AI provider -- run locally with Ollama for full privacy, or connect to Anthropic or OpenAI APIs. Configure in the Settings page or .env.
Install core-skills in 60 seconds and let every conversation build your team's second brain.