Skip to content

SKILLS_GUIDE.md

Latest commit

 

History

History
127 lines (89 loc) · 3.32 KB

SKILLS_GUIDE.md

File metadata and controls

127 lines (89 loc) · 3.32 KB

Agent Skills Guide

VT Code supports repository, user, admin, and bundled system skills using the open Agent Skills SKILL.md format.

Discovery

VT Code discovers skills in this order:

  1. Nearest ancestor .agents/skills from the current working directory up to the git repository root
  2. ~/.agents/skills
  3. /etc/codex/skills
  4. Bundled system skills

If multiple skills share the same name, the nearest repository skill wins, then user, then admin, then system.

VT Code also honors disabled entries from ~/.codex/config.toml:

[[skills.config]]
path = "/path/to/skill/SKILL.md"
enabled = false

Skill Structure

Each skill is a directory containing a required SKILL.md file.

my-skill/
├── SKILL.md
├── scripts/
├── references/
└── assets/

scripts/, references/, and assets/ are optional.

SKILL.md

VT Code accepts the core Agent Skills frontmatter fields plus the client-side disable-model-invocation flag used to hide a skill from the model-facing startup catalog while keeping it available for explicit harness activation:

---
name: my-skill
description: Explain what this skill does and when to use it.
license: Apache-2.0
compatibility: Requires git and network access
allowed-tools: Read Write Bash
metadata:
  owner: platform-team
---

Required fields:

  • name
  • description

Optional fields:

  • license
  • compatibility
  • metadata
  • allowed-tools
  • disable-model-invocation

Legacy VT Code frontmatter such as version, author, when-to-use, when-not-to-use, model, mode, context, agent, network, permissions, container flags, and similar extensions is rejected.

VT Code does not support agents/openai.yaml. That file is Codex-specific and ignored by design.

Prompting Behavior

  • Explicit mention wins: Use the my-skill skill
  • Implicit matching uses description
  • Full SKILL.md bodies are loaded only when a skill is selected
  • Referenced resources are loaded on demand

Commands

List skills:

vtcode skills list

Inspect one skill:

vtcode skills info my-skill

Create a new skill scaffold:

vtcode skills create my-skill

Validate a skill:

vtcode skills validate ./.agents/skills/my-skill

Show configured paths:

vtcode skills config

Slash Command Skills

VT Code also exposes the interactive slash-command surface as skills.

  • Canonical skill names use the cmd-<slash-name> form, for example cmd-status or cmd-review.
  • The /status or /review slash command remains the primary interactive alias.
  • Built-in session/UI commands are surfaced as built-in command skills.
  • Prompt-oriented slash commands are shipped as bundled system skills in the release binary and installed under the system skill cache at runtime.
  • Command skills are intentionally excluded from the default prompt-side skill index to avoid spending context on slash-command metadata that is already exposed elsewhere in the harness.
  • Built-in command skills support info and use, but not load.

Notes

  • vtcode skills create generates a spec-first SKILL.md scaffold with optional commented guidance for disable-model-invocation.
  • User-facing skill metadata in VT Code is limited to the strict SKILL.md fields above.
  • Bundled system skills are surfaced as system scope.