claude-mem

INTERNAL: Task-only worker — invoked via the Agent/Task tool by the graphify-ingest and graphify-update skills, not as a slash command. Knowledge-graph extraction worker for the graphify-ingest and graphify-update skills. Reads a chunk of code/doc/image files, extracts entities + edges + hyperedges per the graphify schema, writes JSON to a specified chunk file. Designed to be dispatched in parallel (one subagent per chunk) for fast semantic extraction at modest cost. Returns a one-line summary. Dispatched by graphify-ingest (initial build) and graphify-update (incremental — only changed docs/images). <example>Context: graphify-ingest skill is processing 9 chunks for an initial build assistant: "Dispatching 9 graphify-extract subagents in parallel — 1 for docs, 8 for code in deep mode." </example> <example>Context: graphify-update detected 3 changed doc files assistant: "Dispatching 1 graphify-extract subagent for the changed docs (code uses cached AST)." </example>

INTERNAL: Task-only worker — invoked via the Agent/Task tool by the project-profile skill, not as a slash command. Project-config scanner for the project-profile skill. Reads a target directory's build/test/lint configuration files (package.json, pyproject.toml, Makefile, CI configs, linter configs) and returns a structured markdown summary of detected commands, tools, and conventions. Used by /project-profile to extract the auto-discoverable parts of AGENTS.md without polluting the main agent's context with raw config-file contents. <example>Context: /project-profile first-run on an unfamiliar repo assistant: "Dispatching mechanical-scanner to detect build/test/lint commands." </example> <example>Context: /project-profile --refresh after CI config changes assistant: "Dispatching mechanical-scanner to re-scan; will diff against existing AGENTS.md." </example>

INTERNAL: Task-only worker — invoked via the Agent/Task tool by the wiki-ingest skill, not as a slash command. Parallel batch ingestion agent for the Obsidian wiki vault. Dispatched when multiple sources need to be ingested simultaneously. Processes one source fully (read, extract, file entities and concepts, update index) then reports what was created and updated. Use when the user says "ingest all", "batch ingest", or provides multiple files at once. <example>Context: User drops 5 transcript files into .raw/ and says "ingest all of these" assistant: "I'll dispatch parallel agents to process all 5 sources simultaneously." </example> <example>Context: User says "process everything in .raw/ that hasn't been ingested yet" assistant: "I'll use wiki-ingest agents to handle each source in parallel." </example>

INTERNAL: Task-only worker — invoked via the Agent/Task tool by the wiki-lint skill, not as a slash command. Comprehensive wiki health check agent. Scans for orphan pages, dead links, stale claims, missing cross-references, frontmatter gaps, and empty sections. Generates a structured lint report. Dispatched when the user says "lint the wiki", "health check", "wiki audit", or "clean up". <example>Context: User says "lint the wiki" after 15 ingests assistant: "I'll dispatch the wiki-lint agent for a full health check." </example> <example>Context: User says "find all orphan pages" assistant: "I'll use the wiki-lint agent to scan for pages with no inbound links." </example>

Autonomous iterative research loop. Takes a topic, runs web searches, fetches sources, synthesizes findings, and files everything into the wiki as structured pages. Based on Karpathy's autoresearch pattern: program.md configures objectives and constraints, the loop runs until depth is reached, output goes directly into the knowledge base. Triggers on: "/autoresearch", "autoresearch", "research [topic]", "deep dive into [topic]", "investigate [topic]", "find everything about [topic]", "research and file", "go research", "build a wiki on".

Plain-language explanation of one node in the project's graphify code graph: source file, type, community membership, and all incoming/outgoing edges with relations and confidence tags. Cheap (no LLM cost — direct graph lookup). Use for "what is X", "explain this function", "show me what touches Y" when Y is a single named symbol. Triggers on: "/graphify-explain", "explain node", "what is X (in the graph)", "describe node", "neighbors of", "what touches X", "show me X's connections" (when about code).

Build a queryable code-structure graph for a codebase using graphify (AST + optional semantic extraction), then file the result into the claude-mem wiki at wiki/code/. Trigger phrases - graphify this codebase, build the code graph, ingest the code, /graphify-ingest, refresh the graph.

Find the shortest path between two named nodes in the project's graphify code graph. Returns the hop sequence with edge relations and confidence tags. Cheap (no LLM cost — NetworkX shortest path on the existing graph). Use for "how does A reach B", "trace the dependency chain from X to Y", "what connects A and B". Triggers on: "/graphify-path", "shortest path between", "how does X reach Y", "trace path from", "dependency chain from A to B", "path A to B" (when about code).

Free-form code-structural query against the project's graphify code graph. BFS (default) or DFS traversal from term-matched start nodes, returning relevant subgraph as node + edge listings with source file pointers and community tags. Cheap (no LLM cost — pure graph traversal). Use for "what's connected to X", "what touches Y", "explore the auth area", or any open-ended code-structural exploration when a graphify graph exists for the project. Triggers on: "/graphify-query", "graph query", "query the graph", "explore the graph", "what's connected to", "what touches", "what's in the X area" (when about code).

3 hooks across 2 events