AI skills for building software factories. My personal library of domain-agnostic agent skills, reused across every project. Small, composable, and hackable — works with any harness that supports skills: Claude Code, Codex, opencode, Cursor, duet, and 70+ others.
npx skills add dzhng/skillsAdd --list to pick individual skills, or copy any skills/<category>/<name>/
folder into your harness's skills directory (e.g. .claude/skills/).
Software is moving from tasks to factories: agents that pursue a goal autonomously until the output can be trusted. The hard part isn't breaking the goal into tasks — it's breaking it into independently verifiable pieces, and knowing where the pieces even are.
These skills run that loop. Treat the unknown as fog of war: map the terrain, carve it into territories that build and verify in isolation, and recursively re-slice whatever hides more map. And re-planning doesn't stop when planning ends — the spec is a living document, updated and re-sliced mid-implementation whenever the work teaches the agent that the plan is stale. Every piece must prove itself — architecture review, code review, and visual review against a baseline — before the loop moves on. Each iteration gets less wrong, until the goal is done.
Proof: one unattended Codex run pursuing a single goal for 1d 16h on top of these skills, slicing and iterating until done.
-
Plan. Ask your agent to
/feature-slicingthe goal. It interviews you, researches the unknowns, and materializes a spec underspecs/<feature>/— a slice graph where every slice is independently verifiable. -
Build. Kick off the loop:
/goal /implement-spec specs/<feature>Add whatever framing fits:
on the xyz branch, orusing /codex as the implementer while you stay the parent orchestrator and reviewer. -
The rest fires on its own. The spec tells the loop when to call the other skills —
/refactor-cleanand a review pass at the end of every slice,/screenshot-critiqueand/compare-screenshotson anything visual,/close-specwhen the last slice lands — and to update and re-slice the plan whenever implementation proves it stale. Every skill is also independently useful: invoke any of them manually whenever you want.
| Skill | What it does |
|---|---|
| feature-slicing | Break a large feature into independently verifiable, human-reviewable slices with API seams and playable checkpoints. |
| implement-spec | Build an existing spec to completion, one reviewable pass at a time, delegating independent slices in parallel. |
| close-spec | Archive a shipped spec and rewrite it from a build plan into a durable rationale record that points back at the code. |
| refactor-clean | Refactor by moving ownership to one clean concept instead of layering compatibility sediment beside the problem. |
| write-docs | Write docs as a glossary of principles and pointers, never a mirror of the code that will rot. |
| codex | Use the local Codex CLI as an independent second agent for review and (on explicit ask) delegated implementation. |
| Skill | What it does |
|---|---|
| compare-screenshots | Judge which image is less wrong against a target you establish — telemetry to locate divergence, not a baseline match. Ships a reusable diff script. |
| screenshot-critique | Use an unprimed subagent as a second set of eyes on visual work before accepting it. |
| preview-shots | Open a curated set of image shots in one macOS Preview window for the user to eyeball. |
| Skill | What it does |
|---|---|
| write-skills | Create or revise agent skills: triggers, leading words, progressive disclosure, and the failure modes to prune. |
| eval-skills | Eval a skill against golden cases — blind runs in fresh subagents, a separate judge, and gap-driven edits. |
| Skill | What it does |
|---|---|
| renderer | Build, debug, or review WebGPU renderer work — three.js/TSL scene layers, node materials, WGSL passes, depth semantics, and browser-verified visuals. |
MIT