Arcanea Skill System

Skills are discrete, reusable capabilities that agents invoke. Each skill has clear inputs, outputs, and purpose.

Skill Categories

Skill Definition Format

Skills are defined in .arc format:

@skill write.narrative
@version 1.0.0
@category creative
@mode create

@description
  Creates flowing narrative content with emotional resonance.

@parameters
  type: story | poem | script | article
  tone: string        # emotional quality
  length: number      # word count target
  structure: string   # narrative arc type

@output
  format: markdown
  metadata: [word_count, reading_time, themes]

@example
  @invoke write.narrative
  @with {
    type: "short_story",
    tone: "melancholic",
    length: 2000
  }

Core Skills

Creative Skills

write.narrative

@invoke write.narrative
@with { type: "story", tone: "hopeful", length: 1500 }

design.concept

@invoke design.concept
@with { subject: "character", style: "minimalist", variations: 3 }

create.character

@invoke create.character
@with {
  archetype: "mentor",
  traits: ["wise", "scarred", "hopeful"],
  role: "supporting"
}

Analytical Skills

analyze.coherence

@invoke analyze.coherence
@with {
  target: "./chapter-3.md",
  aspects: [logic, timeline, character],
  strictness: "standard"
}

research.thematic

@invoke research.thematic
@with {
  content: "./manuscript.md",
  depth: "deep",
  framework: "psychological"
}

Structural Skills

design.structure

@invoke design.structure
@with {
  content_type: "narrative",
  scale: "book",
  requirements: ["three-act", "dual-timeline"]
}

plan.timeline

@invoke plan.timeline
@with {
  project: "novel-draft",
  phases: ["outline", "draft", "revision", "polish"],
  milestones: ["chapter-complete", "beta-ready"]
}

Expressive Skills

write.dialogue

@invoke write.dialogue
@with {
  characters: ["protagonist", "antagonist"],
  situation: "confrontation",
  subtext: "hidden respect"
}

translate.complex

@invoke translate.complex
@with {
  content: "./technical-doc.md",
  audience: "general",
  preserve: ["accuracy", "core-meaning"]
}

Transformative Skills

transform.style

@invoke transform.style
@with {
  content: "./draft.md",
  target_style: "hemingway",
  degree: "moderate"
}

evolve.concept

@invoke evolve.concept
@with {
  seed: "time-traveling detective",
  direction: "darker",
  iterations: 5
}

Skill Composition

Skills can be combined into workflows:

@workflow character-creation

  @stage 1
    @invoke create.character
    @output → raw_character

  @stage 2
    @invoke analyze.coherence
    @input raw_character
    @output → validated

  @stage 3
    @invoke write.narrative
    @with { type: "backstory" }
    @input validated
    @output → final

@return final

Triggers

Skills can be auto-invoked by triggers:

@trigger on-file-create
@pattern "*.character.arc"
@invoke analyze.coherence
@with { aspects: [consistency, depth] }

@trigger scheduled
@cron "0 9 * * *"
@invoke research.thematic
@with { content: "./current-project/" }

Creating New Skills

@skill my.custom-skill
@version 1.0.0
@category creative
@mode create

@description
  Brief description of what this skill does.

@parameters
  param1: type    # description
  param2: type    # description

@output
  format: output-type

@implementation
  1. Validate parameters
  2. Execute core logic
  3. Format output
  4. Return result

Skill Registry

Skills are registered in core/skills/ with this structure:

core/skills/
├── creative/
│   ├── write.narrative.arc
│   ├── design.concept.arc
│   └── create.character.arc
├── analytical/
│   ├── analyze.coherence.arc
│   └── research.thematic.arc
├── structural/
│   ├── design.structure.arc
│   └── plan.timeline.arc
├── expressive/
│   ├── write.dialogue.arc
│   └── translate.complex.arc
├── transformative/
│   ├── transform.style.arc
│   └── evolve.concept.arc
└── integration/
    ├── orchestrate.workflow.arc
    └── synthesize.outputs.arc

Best Practices

1. Single purpose - Each skill does one thing well
2. Clear parameters - Document what each input means
3. Composable - Skills should chain naturally
4. Testable - Include example invocations
5. Versioned - Track changes over time

For the full .arc language specification, see spec/arc-language.md.