Reusable SKILL.md workflows give coding assistants a working method they can use across sessions.
Public · 20 skills · Codex, Claude Code and other SKILL.md-compatible agents
Skills Install Format License Last commit skills.sh
Quick start
Install one skill:
Or replace doc-style with another folder name from the list below. In a local checkout of this repository, inspect what you are installing before using it:
Give the agent a task that matches the skill's description. The intended result is that the runtime discovers the skill and the agent follows its procedure. Its name and description frontmatter tell the runtime when the workflow is relevant; behavior across agent runtimes is unverified.
Demo
What it is
This repository contains versioned workflows for Codex, Claude Code, and other skill-compatible coding assistants. Each skill is a small procedure with its own trigger description, constraints, and execution rules.
Why it exists
Prompts for recurring engineering work can drift when repository-cleanup rules, experiment discipline, documentation standards, or product-design processes get rewritten each time. A skill turns those repeated instructions into a repository-owned procedure that can be reviewed, versioned, and reused.
Core features
Available skills
Agent and app workflows. Trace runtimes, drive Electron apps, and coordinate agents.
agent-engineering — trace and evaluate agent runtimes, tool calls, context, compaction, and prompt caching.
app-control — drive an Electron app through its Chromium DevTools port the way its user does: see, click, type, and check.
multi-agent — multi-agent dispatch protocol: one task, one branch, one worktree, and coordinator integration.
Debugging and tests. Reproduce failures, test behavior, and audit tests.
debugging — root cause before any fix: reproduce, check the environment, and test one hypothesis at a time.
tdd — write a failing behavior test, implement the change, and refactor with focused checks.
test-audit — authoring gate for new tests and an audit that removes low-value, implementation-coupled, or duplicate tests.
Planning and experiments. Define acceptance, isolate variables, and design products.
experiment-workflow — variable-isolated experiment workflow with explicit hypotheses and stress tests.
first-principles — engineering simplification workflow: question, delete, simplify, accelerate, automate.
product-design — product architecture, user-flow design, stack selection, and frontend-first execution planning.
reference-driven-product — improve or rebuild a product from a reference app, screenshots, user feedback, or open-source alternatives.
spec — spec-driven development: acceptance criteria, verification trust, and task importance versus difficulty, written before code.
Documentation and contributions. Check claims, avoid repetition, and make contributor requests clear.
contribution-checklist — check a contributor issue or call for participation for a clear ask, a starting point, and mentoring.
doc-style — evidence-first README, specification, and launch-writing standard, with an audit mode for AI self-talk and stale claims.
dry — one home per fact in docs and rule files: pointers, not copies; no filler.
project-docs — the per-project agent docs (rules, feature map, folder map, current state): one job and one home each.
Visual and document workflows. Handle reports, formulas, graphics, and Persian Word documents.
artifact — handling rules for PDFs, HTML reports, slides, and formula rendering.
latex-authoring — writing, structuring, and compiling LaTeX papers and technical reports.
logo — image-first visual assets: generate with the image tool, then keep the chosen artwork as real SVG paths.
word-farsi-docx — produce editable Persian Word .docx files in B Nazanin with Word-native formatting.
System cleanup. Audit disk use and clean agent clutter.
clean-system — audit disk space and clear agent clutter; ask before clearing caches or Docker.
How it works
Each public skill lives in its own folder:
The file begins with YAML frontmatter:
The rest of the file defines the procedure, output contract, and guardrails. Each skill has one job, and its behavior contract is plain Markdown you can inspect.
The debugging skill adapts systematic-debugging from obra/superpowers (MIT, © 2025 Jesse Vincent). The TDD skill cites mattpocock/skills and obra/superpowers for test-driven development; this repository gives each workflow its own instructions and guardrails.
The repository-owned public set is checked locally with:
This checks the tracked skills' frontmatter and keeps the list above aligned with the tracked set. Run it before committing a change to a public skill.
The smallest useful runtime verification is:
Install one skill into a supported runtime.
Give the agent a task that clearly matches the skill description.
Verify that the runtime discovers the skill.
Check the agent's output against the procedure and guardrails in that SKILL.md.
Where it's going
Current state
Available: The skills listed above are each installable on their own.
Under acceptance: Behavior across agent runtimes is unverified.
Planned: Compatibility checks across multiple supported agent runtimes; more domain-specific engineering workflows when there is a repeated real use case.
Intentionally unsupported: Personal machine-specific configuration, private skills, and third-party skills copied into this repository without a clear reason to redistribute them.
Summary
The repository keeps reviewable, versioned procedures for recurring engineering work in SKILL.md files, so coding assistants can reuse a working method across sessions.