Skip to content

datashaman/code-skills

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

82 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

code-skills

A collection of skills for software engineering workflows.

Installation

Install any skill from this repo using skills:

npx skills add https://github.com/datashaman/code-skills --skill <skillname>

For example:

npx skills add https://github.com/datashaman/code-skills --skill audit-codebase

Skills

/audit-ai-strategy

Audit a codebase's AI strategy through John Cutler's four-bucket lens: bad ideas amplified, good ideas supercharged, genuinely new possibilities, and the meta-skill of reading context. Surfaces where AI is bolted onto broken patterns, where it amplifies what already works, and where the codebase could embrace workflows that only exist because AI is in the loop. Produces a report tagged by bucket with a kill / sharpen / invent / document move per finding, plus an outside-the-box shortlist of small reversible experiments.

Arguments:

  • path (optional): Subdirectory or subsystem to scope the audit to (default: entire repo)
  • focus (optional): Specific concern (e.g. executors, approval flow, PR review, docs)

Usage:

/audit-ai-strategy
/audit-ai-strategy path=src/agents focus="approval flow"

/audit-codebase

Git-based codebase health audit. Runs five diagnostic git commands to identify churn hotspots, bus factor risks, bug clusters, project momentum, and firefighting patterns — then cross-references findings to surface danger zones.

Arguments:

  • path (optional): Subdirectory to scope the analysis to (default: entire repo)
  • since (optional): Time window for analysis (default: "1 year ago")

Usage:

/audit-codebase
/audit-codebase path=src/api since="6 months ago"

/audit-design

Design and WCAG accessibility audit for web UIs. For URLs, uses a live browser to render the page fully before scanning — handles SPAs and dynamically-injected content. For local directories, runs a static scan. Covers contrast ratios, color-only state signaling, semantic HTML structure (headings, landmarks, form labels, alt text, lang), microstandards (OpenGraph, JSON-LD), Tailwind clusters, component health (divitis, clickable divs, inline styles, oversize JSX/TSX, repeated DOM structures), design hygiene (palette, typography, spacing scale, border-radius distribution), AI-slop patterns (purple-violet gradients, 3-col feature grids, emoji in headings, placeholder copy), and a W3C HTML+CSS validator summary. Returns a 0–100 score with actionable findings.

Arguments:

  • url (optional): Deployed URL to audit
  • path (optional): Local directory of source
  • At least one of url or path is required; pass both for plan-vs-implementation divergence mode

Usage:

/audit-design url=https://example.com
/audit-design path=./src
/audit-design url=https://example.com path=./src

/audit-context

Audit your Claude Code setup for token waste and context bloat. Starts from /context output, then audits MCP servers (user-configured and built-in claude.ai *), CLAUDE.md rules and @imports, skills, agents, slash commands, plugins, hooks, all five settings scopes (including managed / enterprise policy), and file permissions. Flags user-configured MCP servers that have well-known CLI equivalents (github→gh, aws→aws, kubernetes→kubectl, etc.) since a CLI costs zero tokens when idle. Mines session JSONL transcripts for behavioral signals (cache hit rate, autocompact frequency, turn-cost percentiles, per-tool error rates, unused skills/agents, repeated Read paths, large tool-result outliers). Cross-references MCP connection logs to catch broken servers that load schemas but never connect. Returns a health score with specific fixes.

Usage:

/audit-context

/audit-jira

Jira project health audit. Queries five dimensions of work item data to identify velocity trends, backlog rot, bug clusters, assignee concentration risk, and firefighting patterns — then cross-references findings to surface danger zones.

Prerequisites: Atlassian MCP connected and authenticated.

Arguments:

  • project (required): Jira project key (e.g. ENG, PLAT) — can be passed positionally
  • since (optional): JQL date expression for the analysis window (default: -52w)

Usage:

/audit-jira ENG
/audit-jira ENG since=-26w

/audit-trello

Trello board health audit. Queries five dimensions of card data to identify throughput trends, backlog rot, blocker clusters, member concentration risk, and overdue patterns — then cross-references findings to surface danger zones.

Prerequisites: trello CLI installed and authenticated (trello auth). Install from https://github.com/mheap/trello-cli.

Arguments:

  • board (required): Trello board name or ID — can be passed positionally
  • done (optional): Name of the Done list (default: auto-detected)
  • since (optional): Cutoff date for age-based analysis (default: 1 year ago)

Usage:

/audit-trello "My Board"
/audit-trello board="My Board" since=2024-07-01

/github-workflow-configurator

⚠️ WIP — not ready for use. This skill is under active development and is not yet stable. Don't install it on a setup you care about.

Interview-driven GitHub workflow setup for agent-assisted delivery. Helps choose and configure team composition, solo-vs-peer product review, architecture review gates, comment-triggered automation, GitHub Actions events, branch/label policy, and automation authority. Produces repo-specific config/docs/templates such as .github/sdlc-agent-workflow.yml, Actions event routing, issue templates, and PR templates.

Usage:

/github-workflow-configurator configure this repo for solo PO workflow
/github-workflow-configurator set up comment-triggered agent automation
/github-workflow-configurator design a GitHub Actions approach for PO + Architect + remote agents

/onepager

Generate a self-contained single-file HTML demo or presentation using Alpine.js + Tailwind via CDN. Emailable, double-clickable, no build step. Aimed at business and marketing users putting an interactive thing in front of a stakeholder without booking engineering — the agent makes all edits, the user only opens the file in a browser.

Ships with four reference patterns the skill draws on: a baseline list-with-details demo, a KPI dashboard with a coaching-mark tour (spotlight overlay, $persist-backed "tour seen" state), a keyboard-driven slide deck with five layouts (cover, bullets, metric, quote, code), and a simulated-AI inbox with thinking-step trace and word-by-word streaming.

Arguments: None — drop a topic in natural language and the skill produces the file.

Usage:

/onepager a demo of our pricing tiers
/onepager Q1 business review slide deck
/onepager walkthrough of our onboarding flow with coaching marks
/onepager a simulated AI assistant for support triage

/harness

⚠️ WIP — not ready for use. This skill is under active development and is not yet stable. Don't install it on a setup you care about.

Control surface for a harness-engineering Claude Code setup at user or project scope. Sub-actions: install (operating-contract CLAUDE.md, four guardrail hooks, /verify and /plan slash commands, auto-memory seeds, settings.json patch); uninstall (symmetric reversal with content-match protection for customised files); update (refresh installed files vs current templates with diffable --merge mode); doctor (end-to-end diagnostic — perms, hook smoke-test, settings JSON validity); adopt (retrofit into an existing project — detects stack, writes a starter scripts/harness-check.sh pass/fail gate, prints next-step install command); snapshot (sanitised mirror of ~/.claude/ to a private git repo); status (report installed / modified / missing per surface); audit (prepare a monthly remote routine that PRs deltas against the latest Anthropic releases and Claude Code community patterns). All idempotent.

Builds on OpenAI's harness-engineering article, Martin Fowler's writeup, and patterns from Boris Cherny, Simon Willison, Jesse Vincent (Superpowers), Geoffrey Huntley, Hamel Husain, and Steve Yegge. Sibling to datashaman/harness-template (the project-scope counterpart).

Arguments: None — the skill detects intent from natural language ("install", "uninstall", "update", "doctor", "adopt", "snapshot", "status", "audit"). Just /harness runs status.

Usage:

/harness                    # status, then ask
/harness install
/harness uninstall
/harness uninstall --all    # also remove CLAUDE.md, memory, env var
/harness update             # refresh files; --merge for diffable side-by-side
/harness doctor             # end-to-end diagnostic
/harness adopt              # retrofit into existing project (scaffolds scripts/harness-check.sh)
/harness snapshot
/harness status
/harness audit              # prep a monthly remote routine

Other skills I like

Evaluates a codebase against five principles that determine AI agent effectiveness, then delivers specific improvement guidance tailored to the project's stack.

The five principles: 100% test coverage, thoughtful file structure, end-to-end types, fast/ephemeral/concurrent dev environments, and automated enforcement.

Modes:

  • audit: Score an existing codebase across all five principles
  • guide: Targeted improvement steps for a specific principle or new project setup

Structure

skills/
  <skill-name>/
    SKILL.md        # Skill definition and instructions

Adding a skill

Create a new directory under skills/ with a SKILL.md file. The frontmatter should include name and description fields. See existing skills for examples.

License

MIT — see LICENSE.

About

A collection of skills for software engineering workflows

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors