avad-help

Files (2)

• SKILL.md
• SKILL.md.tmpl

SKILL.md (first lines)

<!-- AUTO-GENERATED from SKILL.md.tmpl — do not edit directly -->
<!-- Regenerate: bun run gen:skill-docs -->

# /avad-help — Skill Hub & Update Check

When invoked, show all available avadbot skills, help the user pick the right one, and check for updates.

## Step 1: Version Check

Check for updates by comparing the installed version against the latest release:

```bash
# Get installed version from plugin.json
PLUGIN_DIR=$(find ~/.claude/plugins/cache -path "*/avadbot*/.claude-plugin/plugin.json" 2>/dev/null | head -1)
INSTALLED=$(cat "$PLUGIN_DIR" 2>/dev/null | grep '"version"' | head -1 | sed 's/.*"version": *"\([^"]*\)".*/\1/' || echo "unknown")

# Check latest version from GitHub
LATEST=$(gh api repos/agwacom/avadbot/releases/latest --jq '.tag_name' 2>/dev/null | sed 's/^v//' || echo "")
# Fallback: check package.json on main branch
if [ -z "$LATEST" ]; then
  LATEST=$(gh api repos/agwacom/avadbot/contents/package.json --jq '.content' 2>/dev/null | base64 -d | grep '"version"' | head -1 | sed 's/.*"version": *"\([^"]*\)".*/\1/' || echo "unknown")
fi
```

**If update available:** Show a notice at the top:
```
** Update available: v{INSTALLED} → v{LATEST}
   Run: claude plugin update avadbot
```

**If up to date or check fails:** Skip silently.

## Step 2: Show Skill Catalog

Display all available skills in a table:

```
avadbot v{VERSION} — Skill Catalog

Skill                        What it does                              When to use
─────────────────────────── ──────────────────────────────────────── ──────────────────────────────────
/avadbot:avad-plan-ceo-review  CEO/founder plan review — rethink the    Before building: challenge the
                               problem, find the 10-star product        problem, expand/reduce scope

/avadbot:avad-plan-eng-review  Eng manager plan review — architecture,  Before building: lock in the
                               data flow, edge cases, test coverage     execution plan

/avadbot:avad-review           Pre-landing code review with Fix-First   Before shipping: review your
                               — AUTO-FIX mechanical issues             diff, fix issues

/avadbot:avad-ship             One-command ship — validate, sync,       When ready to ship: push +
                               test, review, push, create PR            create PR automatically

/avadbot:avad-qa               QA test a web app — find and fix bugs,   After deploying: systematic
                               generate regression tests                testing + bug fixing

/avadbot:avad-debug            Systematic root-cause debugger —         When stuck on a bug: traces
                               Iron Law, no fix without root cause      data flow, tests hypotheses

/avadbot:avad-document-release Post-ship doc updater — syncs README,    After shipping: update all
                               CHANGELOG, CLAUDE.md with what shipped   docs to match reality

/avadbot:avad-retro            Weekly engineering retrospective —       End of week: commit analysis,
                               commit analysis, team breakdown          patterns, improvements

/avadbot:avad-browse            Headless browser — navigate, click,      When you need to see a page:
                               screenshot, test forms, import cookies   QA, dogfooding, verification

/avadbot:avad-plan-design-review Designer's eye plan review — rates 8     Before building UI: design
                               design dimensions, produces fix plan     critique and score

/avadbot:avad-design-consultation Design system creation — competitive   Starting a project: create
                               research, DESIGN.md, color/font system   DESIGN.md with design system

/avadbot:avad-design-review    Visual QA + fix loop — spacing, slop,   After building UI: find and
                               interaction issues, risk calculator      fix visual issues

/avadbot:avad-careful          Destructive command guardrails — warns   Safety: prevent accidental
                               before rm -rf, DROP TABLE, force-push   destructive commands

/avadbot:avad-freeze           Restrict edits to a directory — locks   Focus: prevent changes outside
                               Edit/Write to a specific path           the directory you're working in

/avadbot:avad-unfreeze         Clear freeze boundary                   After /avad-freeze: unlock edits

/avadbot:avad-guard            Combined careful + freeze in one cmd    Safety + focus: both guardrails

/avadbot:avad-new              Project scaffolding — asks 5 product      Starting a project: scaffold
                               questions, generates 16 pre-filled       16 pre-filled files
                               files, detects stack, creates repo

/avadbot:avad-help              Skill hub — shows all skills,            When unsure which skill to use:
                               context-aware routing, update check      see what's available

/avadbot:avad-hello            Test greeting — confirms plugin works    Testing: verify installation
```

## Step 3: Context-Aware Recommendation

If the user described what they want to do (either in the invocation or as a follow-up), recommend the best skill:

**Role-based routing:**
- "CEO", "founder", "product" → `/avadbot:avad-plan-ceo-review`
- "eng", "engineer", "architect", "technical" → `/avadbot:avad-plan-eng-review`
- "review", "diff", "review PR", "code review" → `/avadbot:avad-review`
- "ship", "push", "deploy", "merge", "create PR" → `/avadbot:avad-ship`
- "QA", "test", "bug", "broken" → `/avadbot:avad-qa`
- "debug", "fix", "stuck", "error", "crash" → `/avadbot:avad-debug`
- "docs", "documentation", "README", "changelog" → `/avadbot:avad-document-release`
- "retro", "retrospective", "week", "metrics" → `/avadbot:avad-retro`
- "browse", "browser", "screenshot", "page" → `/avadbot:avad-browse`
- "design plan", "design critique", "design score" → `/avadbot:avad-plan-design-review`
- "design system", "DESIGN.md", "colors", "typography" → `/avadbot:avad-design-consultation`
- "visual QA", "design audit", "spacing", "AI slop" → `/avadbot:avad-design-review`
- "new project", "scaffold", "init", "start a project" → `/avadbot:avad-new`
- "careful", "safety", "prod mode", "guardrails" → `/avadbot:avad-careful`
- "freeze", "restrict edits", "lock directory" → `/avadbot:avad-freeze`
- "unfreeze", "unlock edits" → `/avadbot:avad-unfreeze`
- "guard", "full safety" → `/avadbot:avad-guard`

From agwacom/avadbot, MIT.

More from agwacom/avadbot