Skills
The catalog currently contains 80 skills under skills/. Each skill is self-contained and starts with a SKILL.md file. Optional scripts, references, and assets live inside the same skill directory.
Source: find skills -mindepth 1 -maxdepth 1 -type d, scripts/validate.sh.
vd:guide is the catalog front door when several skills could apply. The ask-to-ship path is on Getting Started.
Core Delivery Flow
Section titled “Core Delivery Flow”| Stage | Skills |
|---|---|
| Route | vd:guide |
| Define | vd:interview (--grill, --wayfinder) |
| Discover | vd:scout, vd:graphify, vd:research |
| Decide | vd:brainstorm |
| Plan | vd:plan (--audit), vd:scenario |
| Execute | vd:cook, vd:fix, vd:debug, vd:codex-workflow |
| Review and ship | vd:code-review (--refactor), vd:miucr, vd:simplify (--aggressive), vd:security, vd:ship, vd:git |
| Iterate | vd:auto-loop, vd:optimize-loop |
| Orchestrate | vd:ultracook |
vd:interview is the alignment skill: default extracts want, --grill walks an existing plan or idea (the grilling primitive other skills compose), --wayfinder charts a multi-session map of decision tickets. One-session deciding after want is confirmed is vd:brainstorm. vd:plan --audit is the clean-context second look (auto on --deep). vd:ultracook composes these skills by name with checkable done-when gates; it does not own their discipline.
Interview And Decision
Section titled “Interview And Decision”| Skill | Question |
|---|---|
vd:interview | What do you actually want? |
vd:interview --grill | Are these decisions the right ones? |
vd:interview --wayfinder | What must be decided, in what order, across sessions? |
vd:brainstorm | How should I approach this (3+ options)? |
vd:research | Which known option should I pick? |
vd:scenario | What could break? |
Review Lenses
Section titled “Review Lenses”| Skill | Question |
|---|---|
vd:code-review | Ready to land? (posts on a PR) |
vd:code-review --refactor | Does this fit the codebase, or is it slop? (local) |
vd:simplify | Can this read easier with behavior frozen? |
vd:simplify --aggressive | What shape should this have had from day one? |
vd:security | What can an attacker do? |
vd:plan --audit | Does this plan hold up against the codebase? |
vd:miucr | Deterministic owned reviewer CLI (miu-cr) |
Browser And Web
Section titled “Browser And Web”Escalation ladder (start local; vd:browser is the cloud fallback):
| Rung | Skill |
|---|---|
| Isolated logged-in browsing | vd:ego-browser |
| Persistent local Chrome / CDP | vd:agent-browser, vd:browser-profile |
| Vendor-free raw-CDP traces | vd:browser-trace |
| CAPTCHA / anti-bot / proxy | vd:browser (Browserbase) |
| Logged-in end-to-end | vd:web-e2e |
| Performance | vd:web-perf |
Design And Media
Section titled “Design And Media”| Area | Skills |
|---|---|
| Web and frontend | vd:uiuxdesign, vd:opendesign, vd:fastreact |
| Interface and storage design | vd:apidesign, vd:dbdesign |
| Media, files, and social | vd:omnimedia, vd:marketing-design, vd:copywriting, vd:unslop, vd:show-off, vd:file-browser, vd:twitter, vd:devlog |
Docs And Diagrams
Section titled “Docs And Diagrams”| Need | Skill |
|---|---|
Canonical internal docs and ADRs; public site via site | vd:docs |
| Cheap in-chat visual for the current topic | vd:show-me |
| ASCII sketch | vd:text-diagram |
| General SVG or raster | vd:diagram |
| Editable whiteboard | vd:excalidraw |
| Polished, accessible HTML/SVG | vd:diagram-design |
Languages And Migration
Section titled “Languages And Migration”| Skill | Covers |
|---|---|
vd:golang | Idiomatic Go; topic notes under skills/golang/references/ |
vd:gostack | Sam Berthe’s Go libraries (lo, oops, do, mo, slog, hot, ro) |
vd:py2go | Python-to-Go migrations with six project-type playbooks |
vd:cli-ts | Production TypeScript CLI architecture and packaging |
Infra And Ops
Section titled “Infra And Ops”| Skill | Covers |
|---|---|
vd:devops | Docker, Kubernetes, Terraform/OpenTofu, cloud platforms, GitHub Actions, GitOps; deploy-verify checks for image/rollout/CI |
vd:aws | Official AWS CLI, identity-first scoping, incident tracing |
vd:cnpg | CloudNativePG operations |
vd:astro-airflow | Remote Astro / Airflow inspection (af, deployment logs) |
vd:dag-factory | YAML DAG authoring (map vs list dialect) |
vd:managing-astro-local-env | Local astro dev lifecycle |
vd:delegating-to-otto | Headless astro otto |
Product CLIs
Section titled “Product CLIs”Owned CLIs documented as manuals (not generic “how to review/code”):
| Skill | Product |
|---|---|
vd:miucr | miu-cr deterministic reviewer |
vd:miudb | miudb |
vd:vd-cli | vd catalog/install CLI |
Personal Ops
Section titled “Personal Ops”Daily-driver utilities, categorized honestly:
vd:superwhisper, vd:gws, vd:jira, vd:issue-invoice, vd:computer-clean, vd:worktree, vd:herd-worktree, vd:herdr, vd:gopass, vd:journal, vd:workbench, vd:braze, vd:smartsheet, vd:voice-agent
Skill Lifecycle
Section titled “Skill Lifecycle”| Skill | Owns |
|---|---|
vd:skill-creator | Authoring a new SKILL.md (description, hard rules, writing principles) |
vd:skill-management | Scaffold, vendor, validate, release (--create delegates authoring) |
vd:skill-evolve | Improve skills that already exist |
vd:skill-audit | Which skills actually get used |
vd:rule-miner | Distil repeated corrections into CLAUDE.md rules |
vd:agent-readiness | Score a repo against a 30-signal rubric and remediate |
Choosing The Right Skill
Section titled “Choosing The Right Skill”Use vd:guide when the next skill is unclear. Use vd:interview when the ask is underspecified, vd:interview --grill when a plan or idea needs its decisions walked, vd:interview --wayfinder when the deciding itself will not fit one session, vd:scout when you need a map of the repo, vd:debug when behavior is failing, vd:brainstorm when the outcome is known but the one-session approach is not, vd:plan when the implementation path is not yet concrete, vd:cook when a plan is ready to execute, vd:git for a single commit/push/PR/merge (gh pr checks exit 8 is pending, not failure), and vd:ship when the work is tested and ready for remote.
Within review: vd:code-review asks whether a change is ready to land and posts inline PR comments; vd:code-review --refactor stays local and asks whether the change fits the codebase; vd:simplify reduces reading complexity with behavior frozen; vd:simplify --aggressive reshapes a working feature into the form it should have had from day one, deleting compatibility paths only after proving them dead. vd:miucr is the deterministic owned reviewer CLI for when the review itself must be reproducible.
Use vd:docs for canonical internal project docs and ADRs, and its site subcommand to create, modernize, validate, and ship a rendered public developer documentation site.
Use vd:show-me for the cheapest in-chat visual of the current topic. Use vd:diagram-design for polished, accessible, self-contained HTML/SVG diagrams. Use vd:text-diagram for ASCII sketches, vd:diagram for general SVG or raster output, and vd:excalidraw for editable whiteboard canvases.
Use vd:superwhisper to search local dictation history, prepare standups, and diagnose recognition errors. Use vd:braze for Braze CLI reads and explicit opt-in changes. Use vd:smartsheet for bounded sheet reads and authorized row updates. Use vd:voice-agent to operate Retell through vac.
vd:jira uses the vanducng/jira-cli fork for native inline local-image comments. Inside Herdr, vd:worktree delegates current-pane naming to vd:herdr after a successful create. vd:worktree also trusts mise configs in new worktrees so cd works; mise install stays in suggestedInstalls. vd:computer-clean also audits Git worktree storage.
A lingering .ck.json without .vd.json is a rename (mv .ck.json .vd.json), not a skill.