mironmax

kg-capture

Knowledge capture rules. Capture mid-conversation, not after — context is cached, so a write costs almost nothing now but saves full re-derivation next session. Good moments to capture (as things happen, not at task end): - Opened files with no component node → a brief node now saves a re-read later - Discovered how two things connect → an edge, while the insight is fresh - Understood why something works a certain way → a note on the existing node - 10+ min debugging resolved → root cause node before moving on - User expressed a preference, style, or constraint → user-level node - User corrected your approach → capture what was missed, not just the fix - Explained something non-obvious → node before it scrolls away - Approach agreed with user → capture the methodology, not just the decision - Architectural decision made → node with rationale in notes - Context window feels deep → a good moment to check for anything unrecorded When reading a file with no component node, consider creating one. Gist = what it ha

kg-core

Knowledge Graph — persistent memory, your twin across sessions. Treat it as primary context before reaching for any other tool. Session start: if kg_read hasn't been called yet, call it before any task work. kg_read(cwd="<project root>") Output has two sections — USER GRAPH and PROJECT GRAPH. On large graphs the result may start with <persisted-output> and show only a preview; the full output is saved to the file path shown — read it with the Read tool to get the complete picture including session_id. Announce "I have recalled KG Memories" once both sections have been read. Connection refused means the server is down — let the user know: `kg-memory start` will bring it back. If kg-memory isn't found, the install script hasn't been run yet: knowledge-graph/install_command.sh registers both kg-memory and kg-visual. Before searching files, docs, or the web — check what's already known. The graph often has the answer, and reading from memory is faster than rediscovering. Writes during conversation are cheap (cont

kg-maintain

Knowledge graph maintenance. Tend the garden — regular, light care keeps it healthy. Not a separate task: woven into every session, every capture. GARDEN RHYTHM — three modes, applied as needed: Water (routine): after each task, glance at 2–3 recently-touched nodes. Are their gists still accurate? Any note worth adding? Prune (when dense): merge duplicate nodes, shorten verbose gists (→ notes), split oversized nodes, remove stale touches, delete edges to removed concepts. Fertilize (on use): when a node proves valuable, connect it to newly-discovered related nodes. One new edge makes a node far more durable. AFTER CAPTURE: when you save a node, immediately ask — - Do any adjacent nodes now need updating? - Is this a duplicate of something existing? Merge if so. - Does this node's gist still fit, or did context shift? REACTIVE TRIGGERS (act immediately, mid-conversation): Uncertainty (spinning wheels, deja vu, about to search) → kg_search first. About to assume something → check KG; if missing, capture the ass

kg-recall

Knowledge recall rules. Active every session, integrated with all task work. After kg_read, scan all node IDs and gists — anything that feels related to the current task is worth reading in full: kg_read(cwd, id). Lean toward reading more rather than less; a wrong guess costs one tool call, missing context costs the whole task. Gists are headlines (WHAT). Notes and touches hold rationale (WHY) — read those when a decision depends on understanding the reasoning, not just the fact. Three tiers — nodes shift as the graph grows: active → id + gist visible in kg_read archived → id only; edges visible as crumb trails orphaned → invisible in kg_read; reachable via kg_search Following crumbs: an edge pointing to an archived id is an invitation — kg_read(cwd, id) promotes it and surfaces any orphaned neighbors. No edges? Scan the archived list by name. kg_search reaches all tiers when you need to cast a wider net. Before stating an assumption — it's worth a quick kg_search first; the graph may already have the answer.