Build new agent skills. Use when creating diagnostic frameworks, CLI tools, or data-driven generators that follow the established skill patterns.

Skill Builder: Meta Skill for Creating Skills You help create new agent skills that follow established patterns. Your role is to guide skill design, generate scaffolding, and validate completeness. Core Principle Skills are diagnostic frameworks with tools, not feature checklists. A skill diagnoses a problem space, identifies states, and provides interventions. Scripts provide randomization and structure; the LLM provides judgment. Each does what it's best at. Skill Anatomy Every skill has these components: SKILL.md Structure Skill Types Type D: Diagnostic Skills Purpose: Identify problems, recommend interventions Pattern: States → Questions → Interventions Examples: story sense, worldbuilding, conlang Key characteristics: Problem states with symptoms/questions Cross references to intervention tools "What you do NOT do" section enforces boundaries Integration tables mapping to other skills Type G: Generator Skills Purpose: Produce structured output from parameters Pattern: Parameters → Generation → Output Examples: Functions in story sense, phonology in conlang Key characteristics: Input parameters with defaults Randomization with optional seeding Multiple output formats (human, JSON, brief) Quality levels (starter → comprehensive) Type U: Utility Skills Purpose: Support other skills, build infrastructure Pattern: Input → Analysis/Transformation → Report Examples: list builder, skill builder Key characteristics: Meta level operation Quality metrics and validation Templates and scaffolding Cross skill applicability Type O: Orchestrator Skills Purpose: Coordinate multiple skills into autonomous workflows Pattern: Input → Multi Pass Evaluation Loop → Polished Output Examples: chapter drafter Key characteristics: Invokes multiple sub skills sequentially Iterates until quality thresholds met Accumulates context across work units Operates autonomously without human checkpoints Required frontmatter: See skills/fiction/orchestrators/README.md for architectural details. Skill Maturity Scoring (24 points) Skills are evaluated on a 24 point scale parallel to the framework 24 point system. Completeness (11 points) Check Points Criteria Core Principle 1 Bold statement capturing diagnostic essence States 2 3 7 states for diagnostic skills (N/A for generator/utility) State Components 2 Symptoms, Key Questions, Interventions for each state Diagnostic Process 1 Step by step process documented Anti Patterns 2 3+ anti patterns with Problem/Fix structure Examples 2 2+ worked examples showing skill application Boundaries 1 "What You Do NOT Do" section Quality (5 points) Check Points Criteria Self Contained 1 Can be used without reading other skills Type+Mode Declared 1 Required frontmatter fields present State Naming 1 Consistent state prefix matching skill abbreviation Integration Map 1 Documents connections to other skills Tools Documented 1 All scripts have usage documentation Usability (4 points) Check Points Criteria Output Persistence 1 Customized (not boilerplate) persistence section Progressive Disclosure 1 Quick reference section for at a glance use Decision Tree 1 Routing logic for common scenarios Actionability 1 Clear next steps for each diagnosis Execution Intelligence (4 points) — NEW Check Points Criteria Reasoning Requirements 1 Specifies when extended thinking benefits the task Execution Strategy 1 Documents sequential vs. parallelizable work Subagent Guidance 1 Identifies when to spawn specialized subagents Context Management 1 Documents token footprint and optimization strategies Maturity Levels Level Score Description Draft 0 8 Missing core elements Developing 9 14 Functional but incomplete Stable 15 20 Production ready Battle Tested 21 24 Has case studies + full execution intelligence Required Metadata Type (Required) Every skill must declare its type in frontmatter: Type Definition Required Sections diagnostic Identifies problems, recommends interventions States, Diagnostic Process, Anti Patterns generator Produces structured output from parameters Parameters, Generation Logic, Output Formats utility Supports other skills, builds infrastructure Process, Templates, Validation orchestrator Coordinates multiple skills into autonomous workflows Orchestration Loop, Pass Criteria, Iteration Limits Mode (Required) Every skill must declare its mode in frontmatter: Mode Definition User Relationship diagnostic Identifies problem states and recommends Agent diagnoses, user decides assistive Guides without producing content Agent asks questions, user creates collaborative Works alongside user Agent produces, user guides evaluative Assesses existing work Agent reviews, user responds application Operates in real time context Agent runs, user participates generative Creates output from parameters Agent produces, user selects Compound modes (e.g., diagnostic+generative ) are allowed when skills perform multiple functions. Optional Metadata State Naming Convention States must follow a consistent naming pattern: Convention: {ABBREV}{NUMBER}: {State Name} Rules: 1. Abbreviation is 1 3 uppercase letters derived from skill name 2. Numbers start at 0 (for "no X exists" states) or 1 3. State names are descriptive, not just numbers 4. Sub states use decimal notation (4.5, 5.75) when inserting between existing states Standard Abbreviations: Skill Abbreviation Example story sense SS State SS1: Concept Without Foundation dialogue D State D1: Identical Voices conlang L State L1: No Language worldbuilding W State W1: Backdrop World revision R State R1: Overwhelmed endings E State E1: Arbitrary Ending character arc CA State CA1: Static Character scene sequencing SQ State SQ1: Scene Only Pacing brainstorming B State B1: Convergent Ideas research RS State RS1: No Research requirements analysis RA State RA1: Vague Requirements system design SD State SD1: No Architecture chapter drafter CD (Orchestrator uses pass scores, not states) New skills should claim an unused abbreviation and document it here. Integration Graph Requirements Every skill must document its connections to other skills. Required Format: Requirements: Minimum 1 inbound OR 1 outbound connection Complementary skills list for context State level specificity (not just skill to skill) Bidirectional documentation (if A references B, B should reference A) Execution Intelligence Requirements Skills should document how they're best executed by Claude Code. Reasoning Requirements Section Document when extended thinking ( ultrathink ) benefits the skill: Why this matters: LLMs have a completion reward bias—they rush toward visible goals. Extended thinking allocates reasoning time before output, improving quality on complex tasks. This aligns with the LLM Process Design Framework principle. Execution Strategy Section Document sequential vs. parallel work: Context Management Section Document token usage and optimization: Anti Pattern Requirements Every skill must document common mistakes. Minimum Requirements: 3 anti patterns for diagnostic skills 2 anti patterns for generator/utility skills Required Structure: Common Anti Pattern Categories: Category Example Names Scope Creep The Kitchen Sink, The Mission Creep Missing Depth The Surface Treatment, The Checklist Wrong Level The Bottom Up Edit, The Premature Optimization User Relationship The Puppet Master, The Passive Recipient Integration The Orphan Skill, The Boundary Ignorer Script Patterns Standard Script Template Argument Parsing Pattern Data Loading Pattern Output Pattern Data File Patterns Simple List (for entropy/randomization) Quality thresholds: Starter: 10 30 items (demo only) Functional: 30 75 items (usable) Production: 75 150 items (ready) Comprehensive: 150+ items (reference quality) Structured Data (for complex generation) Frequency Weighted Data Diagnostic Process for Building Skills When creating a new skill: 1. Identify the Problem Space What problems does this skill diagnose? What are the symptoms a user would notice? How does this connect to existing skills? 2. Define States Create 3 7 distinct states Each state needs: symptoms, key questions, interventions States should be mutually exclusive but cover the space 3. Design Tools What can scripts do better than LLM judgment? Randomization? Structure generation? Validation? What data does the script need? 4. Map Integrations Which other skills does this connect to? What states in other skills lead here? What states here lead elsewhere? 5. Validate Completeness Run validate skill.ts to check: Required frontmatter fields State definitions with all components Script documentation Integration references Available Tools scaffold.ts Generates skill directory structure and template files. validate skill.ts Checks skill completeness and pattern conformance. Anti Patterns The Feature List Skill Problem: Skill is a list of things it can do, not a diagnostic framework. Fix: Restructure around problem states. What are users stuck on? The Kitchen Sink Problem: Skill tries to do too much, covers multiple problem domains. Fix: Split into focused skills. One skill = one diagnostic space. The Script Without Skill Problem: Script exists but no SKILL.md explains when to use it. Fix: Every script belongs to a skill with documented purpose. The Orphan Skill Problem: Skill doesn't reference or get referenced by other skills. Fix: Add integration section. Map state transitions to/from other skills. The Clone Skill Problem: Skill duplicates another skill's states with different names. Fix: Merge or clearly differentiate the problem spaces. Verification (Oracle) This section documents what this skill can reliably verify vs. what requires human judgment. See organization/architecture/context packet architecture.md for background on oracles. What This Skill Can Verify Structural completeness validate skill.ts checks required sections exist (High confidence) State naming conventions Validates {ABBREV}{N}: Name pattern (High confidence) Frontmatter presence Required type/mode fields present (High confidence) Section presence Anti patterns, Integration Graph, Output Persistence exist (High confidence) State component structure Symptoms/Questions/Interventions present per state (Medium confidence) What Requires Human Judgment State quality Are states mutually exclusive and comprehensive? (Semantic) Integration accuracy Do state transitions make sense across skills? (Contextual) Anti pattern usefulness Do they capture real failure modes? (Experiential) Example relevance Do examples match real use cases? (Domain knowledge) Boundary appropriateness Is the scope correct for one skill? (Design judgment) Available Validation Scripts Script Verifies Confidence validate skill.ts 20 point maturity scoring across Completeness/Quality/Usability High for structure, Low for semantics scaffold.ts Generated files match expected structure High Oracle Limitations The validate skill.ts script: Cannot detect duplicated skill coverage (The Clone Skill anti pattern)

skill-builder