rhize-context-manager

This skill should be used when long-running agent sessions need context compression, structured summarization, compaction, token-per-task optimization, or durable handoff summaries that preserve decisions, files, risks, and next actions.

This skill should be used for diagnosing and mitigating context degradation: lost-in-middle failures, context poisoning, context clash, context confusion, attention-pattern issues, and agent performance degradation caused by accumulated or conflicting context.

Systematic context, session, and memory management for Claude Code development sessions: start/resume/close a working session, preserve and restore context, extract durable memories, and keep long sessions lean. Use whenever the user says "start", "begin", "where were we", "restore session", "new session", "save context", "remember this", "context is getting heavy", "I'm confused/lost", or "done/finished/commit" — and proactively when a session exceeds ~25 messages and needs hygiene, or before implementing a feature when impact-mapping and the component registry should be checked first. Based on Google's Context Engineering guide, enhanced with practical automation hooks and slash commands.

This skill should be used to explain or reason about the foundational concepts of context engineering: what context is, the anatomy of a context window, how attention mechanics work, the U-shaped attention curve, why context quality matters more than quantity, and the mental models needed to interpret every other context-engineering decision. Use this for conceptual explanation, onboarding, and background reading. Route operational work to the specialized skills: debugging attention failures goes to context-degradation, token-efficiency work goes to context-optimization, conversation summarization goes to context-compression, and project-shape decisions go to project-development.

This skill should be used for improving context efficiency: context budgeting, observation masking, prefix or KV-cache strategy, partitioning, token-cost reduction, retrieval scoping, and extending effective context capacity without lowering answer quality.

Build or verify a private, deterministic source-bound code context preview for a specific implementation, diagnosis, impact-analysis, or review task. Use when a bounded dependency-aware pack may reduce targeted reads without hiding critical contracts. Do not use for trivial lookups, highly dynamic code, automatic prompt injection, or as a replacement for exact source verification.

Routing and coexistence brain for the Rhize context stack. Use when deciding WHICH context tool to reach for (Headroom, claude-mem, OpenWolf, Serena, CodeGraph, RTK, graphify, Obsidian, or the preview-only memory-context assembler), when two context layers appear to duplicate or conflict (same context injected twice, slow session start, stacked hooks), or when setting up a new repo and choosing which context tooling it should run. Triggers: "which memory tool", "context is duplicated", "session start is slow", "set up context tooling", "where should this knowledge live".

This skill should be used when agent work needs file-backed context: durable scratchpads, tool-output offloading, just-in-time discovery, cross-agent handoff files, filesystem memory, or cleanup policies for context stored outside the prompt.

Govern Graphify graph.json artifacts for a Rhize Neo4j projection. Use when asked to validate or preview graph ingestion, compile the Rhize ontology, inspect source-bound claims, review possible duplicate identities, reverse an identity decision, measure graph hygiene, preview or reconstruct a consequential decision, test graph tenancy or rollback, or prepare a Neo4j canary. Never use Graphify's direct Neo4j export/push for Rhize governed data.

Use for any question about a codebase, its architecture, file relationships, or project content — especially when graphify-out/ exists, where the question should be treated as a graphify query first. Turns any input (code, docs, papers, images, videos) into a persistent knowledge graph with god nodes, community detection, and query/path/explain tools.

Historical design reference for Graphiti concepts. Use only when reviewing the prior Graphiti decision or comparing its temporal-memory ideas with Rhize's governed Neo4j plans. Do not use this skill to install, configure, route to, or claim adoption of Graphiti.

This skill should be used when deciding whether a session learning, correction, or rule deserves persistent storage — and where to put it so it actually fires: always-loaded config, a reference skill, a hook, or nowhere. Covers drop tests, redundancy and contradiction resolution, the retrieval-cue placement test, and rule phrasing that survives contact with a future session. Route memory *architecture* (vector stores, knowledge graphs, temporal validity) to memory-systems, context-consumption audits to context-budget, and session handoff summaries to context-compression.

Assemble, verify, or explicitly purge a private bounded preview across authorized Rhize memory sources while preserving source authority, conflicts, scope, freshness, and provenance. Use when a task needs context from multiple approved memory lanes, or paired memory measurements. This skill never scrapes transcripts, parses procedural prose, executes recalled procedures, writes memory back, or injects automatically.

This skill should be used for persistent semantic memory in agent systems: cross-session knowledge retention, entity tracking, temporal validity, graph or vector retrieval, memory consolidation, and memory benchmark selection. Route file-backed scratchpads to filesystem-context, handoff summaries to context-compression, and token-efficiency tactics to context-optimization.

Operate and reason about the gated skill-refinement pipeline: headroom learn + claude-mem + skill-monitor signals flow into a human-triaged queue, then skill-forge evolve (SkillOpt-Sleep + safety re-gate) refines target skills with auto-promote for gate-passing SKILL.md edits. Use when the user says "harvest learnings", "refine skills", "drain the refinement queue", asks where headroom learn output should go, or when a scheduled refinement pass needs its rules.

This skill should be used for the tool-interface layer of an agent system specifically: writing tool descriptions agents can route on, designing tool schemas and response formats, naming conventions, actionable error recovery messages, MCP server design, tool-set consolidation, and deciding when to add or remove an individual tool. Use this when the unit of work is a single tool or a set of tools. Route project-shape, pipeline architecture, and task-model-fit decisions to project-development; route deciding whether to introduce sub-agents to multi-agent-patterns.