graph-rag
Relationship-aware retrieval using graph traversal, entity anchors, community expansion, and hybrid vector plus graph search. Use when chunk similarity alone misses paths, entities, or subsystem context.
What this skill does
# Graph RAG ## Overview Use graph-native retrieval when the answer depends on relationships, not just similar text. Graph RAG works well for entity-heavy systems, architecture questions, causal chains, and multi-hop queries that plain vector retrieval often misses. ## When to Use - The user asks how two concepts connect - The answer depends on paths, dependencies, or neighborhoods - Important context is split across multiple files or documents - Vector search returns individually relevant chunks but weak overall explanations - You already have entities, references, or graph structure available ## Retrieval Patterns ### Entity Anchor Retrieval Resolve the question to known entities first, then retrieve around them. ### Neighborhood Expansion Expand one or two hops across relevant relations only. ### Path Retrieval Find the path between two anchors when the question is about connection or causality. ### Community Retrieval Pull the subsystem or cluster around the anchor when local context matters more than one edge. ### Hybrid Retrieval Use vector search to find candidate anchors, then use the graph to expand and explain. ## Process ### Step 1: Classify the Question Graph RAG is a fit when the question is one of these: - connection: "how is A related to B?" - path: "how does data get from A to B?" - neighborhood: "what else is involved with A?" - subsystem: "what belongs to this area?" If the question is simple lookup, plain retrieval may be enough. ### Step 2: Resolve Anchors Identify entities, files, symbols, tables, or services named in the question. If anchor resolution is fuzzy: - use semantic search first - rank candidates - keep the confidence visible ### Step 3: Expand With Bounded Traversal Expand only across relations that matter to the question: - imports - calls - references - belongs_to - decided_by - documented_in Bound the retrieval: - max depth - max nodes - relation allowlist ### Step 4: Build Prompt Context Assemble context as structured evidence, not a raw graph dump: ```markdown ## Anchors - AuthController - SessionToken ## Relevant Path AuthController -> AuthService -> TokenStore -> sessions table ## Supporting Evidence - src/auth/controller.ts:42 - src/auth/service.ts:88 - src/data/token-store.ts:21 - docs/decisions/2026-01-15-auth.md:12 ``` ### Step 5: Answer With Relationship Context The answer should explain: - what the relevant nodes are - how they connect - which evidence supports the path - where uncertainty remains ## Selection Guide | Question Shape | Retrieval Strategy | |---|---| | direct lookup | vector or keyword only | | entity + neighbors | anchor + neighborhood expansion | | how A connects to B | anchor + path retrieval | | subsystem overview | anchor + community retrieval | | fuzzy question with named concepts | hybrid vector + graph | ## Common Rationalizations | Rationalization | Reality | |---|---| | "Vector search already found the files" | File relevance is not the same as relationship explanation. | | "Dump the whole graph into the prompt" | Large raw graphs waste context and hide the important path. | | "More hops is better" | Unbounded traversal quickly turns into noise. | ## Verification - [ ] The question actually needs relationship-aware retrieval - [ ] Anchors are resolved with visible confidence - [ ] Traversal is bounded by depth and relation type - [ ] Prompt context contains paths and evidence, not a raw graph dump - [ ] The final answer explains both the conclusion and the connection path ## Anti-Rationalization Table | Excuse | Counter | |--------|---------| | "Vector search already found the files" | File relevance is not the same as relationship explanation. | | "Dump the whole graph into the prompt" | Large raw graphs waste context and hide the important path. | | "More hops is better" | Unbounded traversal quickly turns into noise. Bound the expansion. | | "Graph RAG is overkill for this question" | If the question involves connections, graph retrieval is the right tool. | | "I'll skip anchor resolution and just expand" | Without anchors, expansion is random. Resolve anchors first for targeted retrieval. |
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.