Extraction Skills

Skills tell Multiplist what to look for in your specific content type. A coach creates a "Client Breakthroughs" skill. A developer creates an "Architecture Signals" skill. A researcher creates a "Research Distillation" skill.

Skills are the differentiator — they teach the vault the shape of your work, beyond the default nine extraction categories.

# Anatomy of a skill

Every skill defines:

• Name & description — human-readable.
• Domain — grouping label (research, engineering, design, coaching, etc.).
• Instructions — the full natural-language prompt the extraction engine follows when this skill runs.
• Schema — optional structured output fields. Each field has a name, type, description, and optional enum values.
• Trigger conditions — optional. Conditions that auto-match this skill to a source (keywords, source type, content length, tags).
• Parent skill — optional. Inherit instructions and schema from another skill.

# Workflow

1. Create with create_skill. Define the name, domain, instructions, and schema.
2. Match with match_skills against a given source to see which skills apply.
3. Extract with trigger_extraction passing skillId — the extraction run uses this skill instead of (or alongside) the default categories.
4. Iterate — refine the instructions, rerun against the same source, compare results.

# Example — "Architecture Signals"

A developer who collects AI-conversation transcripts wants to automatically pull out architectural decisions, tradeoffs considered, and hidden invariants.

• Name: Architecture Signals
• Domain: engineering
• Instructions: "Extract every architectural decision with its reasoning. Flag each tradeoff that was considered. Note any invariants that are implied but not stated."
• Schema:

- decision (string) — what was decided - reasoning (string) — why - tradeoffs (array of strings) — alternatives considered - invariant (string, optional) — implicit rule the decision preserves

• Trigger conditions: keywords: ["architecture", "design decision", "tradeoff"]

Every source this skill runs against produces seeds matching that schema, searchable by field.

# Why skills matter

The default nine categories are a good starting point — but meaning is domain-specific. Skills let the vault grow with you. Over time, your library of skills becomes the operational definition of what your work produces.

Browse with list_skills. Read a skill's full details with get_skill.