OpenParable
Skip to content
← Back to skills

Vcg

Apps & Automation
vcg

Vcg is an Agent Skill for apps & automation published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Short, explicit alias for vc-router. Use only when the user explicitly invokes `use vcg:` and wants the shortest reliable router command.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/data/vcg/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/vcg, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

VCG (VC Go Alias)

VC Defaults

  • Prefer fast iteration and shipping a working baseline over perfection.
  • Make safe default choices without pausing; record assumptions briefly.
  • Ask questions only after delivering an initial result, unless the workflow requires confirmation for safety/legal reasons.
  • Keep outputs concise, actionable, and easy to extend.
  • Assume the user is non-technical; avoid long explanations and provide copy/paste steps when actions are required.
  • Treat non-explicit triggers (e.g., "vc go") as normal text; ask the user to rephrase using use vcg:.

VC Fast Path

  • Classify the task in one pass.
  • Select a single best-fit skill; avoid chaining unless required.
  • Execute immediately; collect assumptions and questions for the end.
  • If the task is multi-step or open-ended, default to a finish-style workflow.

Sub-Agent Assist (Optional)

If the request is large/ambiguous and collaboration tools are available, spawn 1–2 sub-agents to parallelize:

  • Repo scan (within scope roots): locate relevant files/config and constraints
  • Risk/validation scan: test strategy, edge cases, safety concerns

Rules:

  • Timebox and keep outputs short.
  • Sub-agents report findings only; main agent decides routing and applies changes.

VC Quick Invoke

  • use vcg: <goal>

Scope Lock (Required)

  • Before any file search, determine scope roots:
    1. If a .vc-scope file exists in the current directory or any parent, use the closest one.
      • Each non-empty, non-comment line is an allowed path.
      • Relative paths are resolved from the .vc-scope file directory.
    2. Else, if inside a git repo, use the repo root.
    3. Else, use the current working directory.
  • Only run rg, find, or any filesystem scans inside the scope roots.
  • Never scan $HOME or / unless the user explicitly asks.

Routing Rules

  • This repo intentionally ships only vc* skills.
  • Finish-to-end requests: route to vc-phase-loop.
  • Planning/execution loops: route to vc-phase-loop.
  • Everything else: route to vc-phase-loop as the safe default.

Execution Rules

  • Read only the selected skill's SKILL.md and follow its defaults.
  • Keep the user flow simple: deliver a first pass, then ask for corrections.
  • If uncertain between two skills, pick the one with narrower scope.
  • If the user did not explicitly invoke use vcg:, ask them to rephrase using use vcg: and stop.