From lynk-semantic-layer
Answer questions about the Lynk semantic layer — what's in `.lynk/` (instance lookups) and Lynk concepts themselves (primitives, file types, placement). Read-only: never edits, never calls the backend. Use this skill for any question about Lynk or `.lynk/`, even simple ones. Lynk distinguishes primitives that general analytics vocabulary blurs — e.g. a *metric* under `metrics:` is not a *feature of type metric* under `features:`. Don't answer from prior knowledge; run this skill so the answer is doc-grounded. Instance triggers: "does the player have a metric for total games played?", "what metrics does X have?", "list features of X", "where is Y defined?", "what's the SQL for X?", "how many entities do I have?". Concept triggers: "what is a knowledge file?", "metric vs. metric feature?", "formula vs. metric feature?", "where should glossary terms go?", "what is a clarification policy?". For edits use `lynk-build`; for quality `lynk-evaluate`; for backend checks `lynk-validate`; for sources `lynk-sources`.
How this skill is triggered — by the user, by Claude, or both
Slash command
/lynk-semantic-layer:lynk-askThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill answers questions about the Lynk semantic layer. It is read-only — it never writes to `.lynk/`, never calls the Lynk API. Two question shapes are in scope:
This skill answers questions about the Lynk semantic layer. It is read-only — it never writes to .lynk/, never calls the Lynk API. Two question shapes are in scope:
.lynk/?": entities, metrics, features, relationships, glossary, context files.Every answer must be grounded in the docs (and, for instance answers, the actual .lynk/ files) — Lynk distinguishes primitives in ways general analytics vocabulary blurs (e.g., a metric under metrics: vs. a feature of type metric under features:). Do not answer from prior knowledge alone.
Always do this first, before classification or .lynk/ reads. Two fetches:
WebFetch https://docs.getlynk.ai/llms.txtWebFetch https://docs.getlynk.ai/concepts.mdThese two anchors and how to walk from the index to leaf pages are the doc-navigation convention written up in references/lynk-docs.md.
This grounds every answer in correct Lynk vocabulary and gives you a map of doc pages to navigate to next. Skipping this step is what causes the most common failure mode for this skill — confidently confusing related primitives (e.g., treating a metric feature as a standalone metric) because general analytics vocabulary doesn't preserve Lynk's distinctions.
| Shape | Examples |
|---|---|
| Instance | "does X have Y?", "what metrics on X?", "where is Y defined?", "list features of X" |
| Concept | "what is a knowledge file?", "metric vs. metric feature?", "where should X go?" |
| Both | "what is a metric feature, and does my player have any?" |
If ambiguous, ask via AskUserQuestion.
WebFetch only the concepts/<concept> or file-types/<type> pages relevant to the question. For placement questions ("what goes in X?", "where should Y live?"), the file-type spec is the canonical answer.find ./.lynk -type f | sort, identify which file(s) own the artifact, and read only those. For an entity question, that's the entity YAML plus its knowledge / task-instructions files.metrics:) or a metric feature (under features: with type: metric). "Knowledge" can mean entity, domain, or business knowledge. A lead like "Yes — there's a metric called X" is wrong if X is actually a metric feature: it plants the wrong primitive in the user's head and reproduces the exact failure this skill exists to prevent. The lead sentence must name the exact primitive — e.g. "Yes — total_games_played is a metric feature on player, not a standalone metric."points_per_game metric on the entity), cite it and call out the gap explicitly: "the concept exists in your glossary as Z, but isn't modeled as a metric/feature/relationship." This is what makes the lynk-build handoff land — the user sees the modeling gap, not just an empty result.file:line paths and quote the relevant YAML or markdown.lynk-build to add it.lynk-evaluate.lynk-validate.lynk-sources.Never edit files from this skill — edits belong in lynk-build, which has its own plan / confirm / write / evaluate flow that this skill should hand off to rather than bypass.
player.yml → metrics:").Searches MemPalace before answering questions about past work, people, projects, or prior decisions. Returns verbatim stored content instead of guessing from model memory.
Guides Payload CMS config (payload.config.ts), collections, fields, hooks, access control, APIs. Debugs validation errors, security, relationships, queries, transactions, hook behavior.
Implements vector databases with Pinecone, Weaviate, Qdrant, Milvus, pgvector for semantic search, RAG, recommendations, and similarity systems. Optimizes embeddings, indexing, and hybrid search.
npx claudepluginhub lynk-ai/build-agent --plugin lynk-semantic-layer