Skip to content
Back to skills

Guide

ASecurity

Catalog front door. Map an ask to the right vd: skill and the next hand-off so agents and humans do not memorize the catalog. Use when the user asks 'which skill', 'how do I start', 'what should I run', 'guide me', 'router', or when several skills could apply (review lenses, browser ladder, diagram picker, get-it-done paths).

  • 8 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 19, 2026
ai-agentsgospringcode-reviewsecurityperformance

Works with

  • cli

Security analysis

A100/100

Scanned October 7, 2026

npx -y skills add vanducng/skills --skill guide --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Guide?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Guide
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vanducng-guide/badge)](https://www.skillsdirectory.com/skills/vanducng-guide)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: guide
description: "Catalog front door. Map an ask to the right vd: skill and the next hand-off so agents and humans do not memorize the catalog. Use when the user asks 'which skill', 'how do I start', 'what should I run', 'guide me', 'router', or when several skills could apply (review lenses, browser ladder, diagram picker, get-it-done paths)."
license: MIT
argument-hint: "[what you are trying to do]"
metadata:
  author: vanducng
  version: "0.1.3"
---

# Guide

This skill picks the next skill. It does not interview, plan, or edit source. Name one skill (plus a flag if needed), say why, and stop unless the user asks you to run it.

## Delivery spine

One path: **interview → brainstorm → plan → cook → ship**. `vd:ultracook` runs the smallest slice of the same path.

| If | Use |
|---|---|
| Who / why / success / out of scope is missing | `vd:interview` |
| A plan or idea needs its decisions walked | `vd:interview --grill` |
| Deciding will not fit one session | `vd:interview --wayfinder` |
| Want is clear, how is not | `vd:brainstorm` |
| Approach is picked | `vd:plan` → `vd:cook` → `vd:code-review` → `vd:ship` |
| Something is broken | `vd:debug` → `vd:fix` |
| Drive the whole thing | `vd:ultracook` |

## Three "just get it done" paths

| Path | Pick when |
|---|---|
| `vd:cook --quick` | Tight, no plan, one surface. You will still verify and test. |
| `vd:fix` | Known bug, narrow blast radius. `--quick` on cook covers similar ground. |
| `vd:ultracook` (direct) | User asked to orchestrate and the task is trivial. Ultracook classifies; it should not open a goal-dir. |

If two apply, prefer `vd:fix` for failures, `vd:cook --quick` for small builds, `vd:ultracook` when the user wants a conductor.

## Review lenses

| Question | Skill |
|---|---|
| Ready to land? | `vd:code-review` (posts on a PR) |
| Does it fit the codebase / is it slop? | `vd:code-review --refactor` (local) |
| Can it read easier with behavior frozen? | `vd:simplify` |
| What shape should this have had from day one? | `vd:simplify --aggressive` |
| What can an attacker do? | `vd:security` |
| Does this plan hold up? | `vd:plan --audit` |
| Deterministic owned reviewer CLI? | `vd:miucr` |

## Browser escalation

| Rung | Skill |
|---|---|
| Logged-in browsing in an isolated ego space | `vd:ego-browser` |
| Persistent local Chrome, CDP, video, network mocks | `vd:agent-browser` (+ `vd:browser-profile` to launch) |
| Vendor-free raw-CDP traces | `vd:browser-trace` |
| CAPTCHA / anti-bot / proxy (Browserbase) | `vd:browser` |
| Logged-in end-to-end flows | `vd:web-e2e` |
| Performance | `vd:web-perf` |

Start local. Escalate to `vd:browser` only when a local run hits a wall.

## Diagram picker

| Need | Skill |
|---|---|
| Explain the current topic in chat | `vd:show-me` |
| ASCII sketch | `vd:text-diagram` |
| General SVG / raster | `vd:diagram` |
| Editable whiteboard | `vd:excalidraw` |
| Polished, accessible HTML/SVG with layout guidance | `vd:diagram-design` |

## Interface

| Need | Skill |
|---|---|
| Design, build, or review a product UI | `vd:uiuxdesign` |
| Exact polish values: radius, shadows, press scale, icon motion | `vd:better-ui` |
| Whether motion should exist, easing, springs, popovers, drag | `vd:design-eng` |

## Discover / decide / iterate

| Need | Skill |
|---|---|
| Map the repo | `vd:scout` |
| Cited comparison of known options | `vd:research` |
| Edge cases / blast radius | `vd:scenario` |
| Keep iterating one command | `vd:auto-loop` |
| Mechanical quality grind | `vd:optimize-loop` |
| Which skill do I run? | this skill |

## Output

1. Restate the ask in one line.
2. Name **one** skill id (and flag).
3. One sentence why the neighbors are wrong.
4. The next hand-off after that skill finishes.

Do not load the target skill's body unless the user says to run it.

## Tie-break with jev (optional, pi only)

When exactly two candidates both plausibly fit and the tables above do not settle it, one `jev` call (if the harness exposes the tool) can break the tie: Choice over the tied candidates, criteria = each skill's "pick when" line, state = the ask restated in one line. Route: go with the winning option at probability ≥0.6; below that, prefer the spine default (`vd:fix` for failures, `vd:cook --quick` for small builds) and say the ask was ambiguous. If jev is unavailable or errors, decide inline - never block on it.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…