spillwavesolutions/project-knowledge-capture
Project Knowledge Capture: meetings, experiments, decisions, TicketLink/WikiTicket work, and project-memory nouns into a durable OKF knowledge graph. Multi-host bindings and write isolation for shared second brains.
Capture an Assumption (weaker than a Decision) linked to Features. Promote to DecisionRecord when validated. Use for working hypotheses and unproven beliefs.
Capture a short decision statement as an OKF DecisionRecord (ADR-style) linked to Features, Designs, Requirements, Meetings, or Experiments. Use for ADRs, architecture choices, and lightweight decision logging.
Capture research, competitive scans, user findings, and discovery notes as OKF Discovery concepts with source, confidence, and informs links. Use for research dumps, market scans, and exploratory findings.
Capture experiment or spike results into an OKF Experiment concept with hypothesis, method, results, conclusion, and typed informs links. Use for spikes, A/B tests, POC write-ups, and engineering experiments.
Capture meeting notes into OKF Meeting concepts, extract DecisionRecords, and optionally create WikiTicket action items + TicketLinks. Use when the user pastes meeting notes, transcripts, or asks to record a meeting decision trail.
Capture a GitHub PR as an OKF CodeChange concept via gh pr view (or JSON fixture). Use when documenting significant merges.
Capture an open Question that may block a Feature until answered. Use for unresolved design questions and research gaps.
Capture Slack/Discord/chat thread paste as Meeting or Discovery. Scrubs secrets/PII. Use when user pastes a conversation.
Ingest Fireflies, Otter, Granola, or plain speaker-labeled transcripts into a Meeting concept. Always scrubs secrets/PII. Use when user pastes or provides a transcript export.
Build a progressive-disclosure context pack for a Feature (or any concept). Supports --tiny (1 hop / 8 nodes) for chat focus and mermaid graphs. Auto-run when starting work on a Feature.
Weekly/daily knowledge digest — recent captures, decisions, open questions, unvalidated assumptions, needs-verification queue, doctor highlights.
Run a one-screen health check on a PKC knowledge bundle — broken links, orphans, thin Features, decision conflicts, open Questions, stale Discoveries/Assumptions, toolchain (ripgrep). Use when auditing knowledge quality or before a release.
Multi-repo federated knowledge roots — list remotes, search across them, write shadow federation index. Remotes are read-only; link with maps_to.
Import MADR or adr-tools Architecture Decision Record directories into OKF DecisionRecords.
Refresh, inspect, or drop the disposable SQLite/FTS5 knowledge index. Use when search/pack/validate is slow, after a large git checkout, or when the user asks about the PKC index.
Scaffold a Project Knowledge Capture OKF bundle with all PKC catalogs (meetings, experiments, discoveries, decisions, features, …), root index.md, and log.md. Use when starting knowledge capture in a repo or creating knowledge/ / .okf for PKC.
Add a typed edge between two OKF concepts (satisfies, implements, designed_by, decides, informs, originates_from, tracks, maps_to, etc.). Use when connecting meetings to decisions, features to requirements, or tickets to work concepts.
Materialize WikiTicket worklog fold and docs/ artifacts into OKF concepts (Features, TicketLinks, Designs, DecisionRecords, Releases, Specs). Use when syncing work items into the knowledge graph or after major planning/docs updates.
Promote a Discovery, Experiment, or Meeting into a formal Requirement, Feature, Specification, Design, or DecisionRecord while preserving originates_from / informs provenance. Use when informal knowledge is ready to become durable product structure.
Generate release notes from Release concepts and Feature/CodeChange edges (released_in, lands_in, implements).
Spawn the knowledge-retriever sub-agent to search and pack project memory without contaminating parent context. Use before answering from the OKF tree, or when starting work that needs Feature/Decision context.
Full-text search over the knowledge bundle (AND terms, type filters). Use when finding decisions, meetings, or features by keyword.
Open or close an isolated second-brain write session (worktree + PR). Use before capturing or materializing when multiple agents share one knowledge remote.
Check PKC toolchain (python, ripgrep, SQLite FTS5) and optionally install ripgrep after the user consents. Use when search/pack is slow, rg is missing, or the user asks to set up PKC.