haac-aikit

One AGENTS.md config, seven AI coding tools see it. Plus four HTML skills that produce single-page artifacts instead of walls of markdown.

Works with Claude Code · Cursor · Copilot · Windsurf · Aider · Gemini CLI · OpenAI Codex CLI.

Install

npx haac-aikit

30-second wizard. Headless: npx haac-aikit --yes --tools=claude.

The four HTML skills

aikit add --html

| Skill | Use it when | Output | |---|---|---| | /docs | Documenting a feature or gotcha | docs/<name>.html | | /decide | Picking between 2-4 technical options | docs/decisions/<date>-<slug>.html | | /directions | Exploring visual design variants | docs/directions/<date>-<slug>.html | | /roadmap | Handing an implementation plan to an implementer | docs/roadmaps/<date>-<slug>.html | | /design (opt-in: aikit add design) | Codifying the project's design language as a contract | DESIGN.md + docs/design/index.html |

Each command writes one self-contained HTML file — no build step, no CDN, commit and share the link. Built on Thariq Shihipar's "Unreasonable Effectiveness of HTML for Claude Code".

Does /design actually help?

We ran the same four prompts twice — once with the skill installed, once freestyle. Same model, same inputs.

| The ask | With /design | Without | |---|---|---| | "Read my homepage HTML, build a DESIGN.md" | 86% | 71% | | "Just synthesize one — here's the vibe" | 86% | 43% | | "Tweak the primary color, leave the rest alone" | 100% | 100% | | "Build it from this screenshot" | 100% | 57% | | Average across all four | 93% | 68% |

+25 points on average. The freestyle AI usually misses three things: marker-bounded sections (so /design refine can't surgically update later), hex codes inside backticks (the showroom's color pickers can't bind to them), and parts of the brief — "no gray" turned into five shades of gray. /design bakes all three in.

What else you get

• Cross-tool fan-out. Skills, agents, hooks, and MCP servers fan out per selected tool in its native format. Pick --tools=cursor and get .cursor/rules/ + .cursor/hooks.json + .cursor/mcp.json. Pick --tools=codex and get .codex/agents/*.toml + .codex/config.toml with MCP servers.
• Safety hooks. Blocks force-push to main, secret commits, rm -rf on sensitive paths, reads of .ssh/ / .aws/ / secrets/.
• Rule observability. Telemetry tracks which rules fire, which are violated, which are dead weight. aikit doctor --rules shows the report.
• Built-in skills + agents. 14 always-on skills (TDD, debugging, brainstorming, planning, ...) + 11 opt-in. Specialist agents for orchestration and PR description.

Commands

aikit                interactive wizard
aikit add --html     install the four HTML skills + templates
aikit add <name>     add a single skill / agent / hook
aikit sync           regenerate per-tool files from .aikitrc.json
aikit doctor --rules sanity check + observability report
aikit report         rule-adherence summary (markdown or json)

Full reference: aikit --help. Detailed audit per tool: AUDIT.md.

Status

Pre-1.0. Expect breaking changes between minor versions. 0.12.0 ships with cross-tool parity for 6 of 7 tools (Aider has no native rule loader; we ship a skills index in CONVENTIONS.md instead). All blockers from the pre-publish audit have been fixed.

Heads up (0.14.0): skills migrated to folder format — .claude/skills/<name>/SKILL.md instead of .claude/skills/<name>.md. Re-run aikit sync after upgrading; see CHANGELOG for migration notes.

• Site: https://hamad-center.github.io/haac-aikit/
• Try / discuss: issue #1

MIT. See ATTRIBUTIONS.md.