Skip to content

Refactor skill system to adopt AgentSkills.io SKILL.md standard #150

@Aaronontheweb

Description

@Aaronontheweb

Goal

Align Netclaw skill ingestion and authoring with the AgentSkills.io specification so skill discovery is portable, deterministic, and context-efficient.

Why

Current skill plumbing is markdown-comment based (e.g. <!-- description: ... -->, <!-- triggers: ... -->) and does not fully match the open standard (SKILL.md + YAML frontmatter). As the skill ecosystem grows, this makes interoperability and routing quality harder.

Standards target

Use the AgentSkills.io spec as baseline:

  • required SKILL.md per skill directory
  • YAML frontmatter with required metadata:
    • name
    • description (includes when-to-use cues)
  • progressive/contextual discovery:
    • load only metadata at startup
    • activate by loading full SKILL.md only when relevant
  • optional standard fields where useful (compatibility, metadata, allowed-tools)

Reference:

Scope proposal

  1. Scanner/registry refactor
    • Discover skills by */SKILL.md convention.
    • Parse YAML frontmatter metadata.
    • Preserve backward compatibility for legacy .md skills during migration window.
  2. Contextual discovery plumbing
    • Keep startup metadata compact.
    • Improve matching/routing signals from standardized frontmatter + descriptions.
  3. Authoring docs + templates
    • Update skill-authoring guidance to frontmatter-first format.
    • Add canonical skill template generated by CLI/workflow.
  4. Validation gates
    • Add validation checks in tests and/or CI for frontmatter schema and name/path consistency.
    • Optionally use skills-ref validate in dev workflow.
  5. Migration plan
    • Convert system skills to SKILL.md format.
    • Define deprecation timeline for legacy comment metadata.

Acceptance criteria

  • Netclaw can discover and index AgentSkills.io-compliant skills with frontmatter.
  • Agent prompt context includes concise metadata only until activation.
  • Skill activation/retrieval behavior remains stable or improves under load.
  • Existing built-in/system skills are migrated or compatibility shim is documented.

Related

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions