What's in the kit

The kit is two independent systems that share one install path. skills/ teaches an AI coding agent how to work — session discipline, model-specific behavior, memory, coordination. coding-standards/ teaches it what correct code looks like for this project — language rules, file-role conventions, framework overrides. Either can be adopted alone; together, the coding-standards skill is what loads and enforces the standards chain during a session, so a project that wants enforcement needs both folders.

Every skill is a single portable markdown file with name + description frontmatter — the description is trigger-rich ("trigger when…") so it loads only when relevant — and reads a project-local, gitignored learnings.md sidecar at start and appends a lesson at end, so this repo stays the clean upstream while each install compounds its own experience. Every standard is testable and declarative, with wrong/right examples wherever a rule could be misread, and flags a gap (RULE AI-12) rather than inventing a rule where the standards are silent. They interlock at one seam: the install-kit skill installs both in one pass, and the coding-standards skill is what actually loads and walks the standards chain during a session — without it, coding-standards/ is just reference documentation.

skills/ — the self-improving skills library

Full catalog: skills/README.md

coding-standards/ — the layered standards system

Full map: coding-standards/index.md

Install into a project

1. Copy skills/ — and coding-standards/ if the project should carry the standards — into the project root.

2. Open Claude Code (or your AI coding tool) in that project and say:

   read skills/README.md and install

3. Choose everything or pick — the installer detects which coding tool you're in, copies the chosen skills to that tool's skill root (.claude/skills/ for Claude Code, .opencode/skills/ for OpenCode), wires the standards into that tool's session file (CLAUDE.md or AGENTS.md respectively), and reports what was installed.

Details, including per-skill manual installs: skills/README.md → Installing into a new project.

Key ideas

• Self-improving skills. Every skill reads a learnings.md sidecar at start and appends one distilled lesson at end. Learnings are per-project (never committed here, never copied by the installer) — each project's copies adapt to that project.
• Adaptive to change. Skills that describe living things (Claude models, libraries, frameworks) carry staleness guards: verify against the live source, follow it over the skill text, log the correction.
• Standards as law, gaps flagged. The standards system is declarative and testable; where it is silent, the AI flags the gap instead of inventing a rule (RULE AI-12).
• Sequential by default, parallel when you say so. The role-session skill coordinates multiple parallel AI sessions — Claude Code, OpenCode, or both (role charters, atomic mkdir file claims, a git commit token, session ids) and switches itself off in projects without the parallel structure.
• Token-lean by design. Skill descriptions are hard-capped, bodies stay under ~120 lines, read-efficiency rules are part of the standards, and the memory-gardener skill prunes accumulated knowledge.
• Intent before implementation, a self-check before handoff. intent-capture pins down goal/constraints/done-when on ambiguous asks before any plan is made; pre-merge-gate re-checks a diff against the loaded standards before a commit or review handoff — both are prose protocols, not tool-specific.
• Claude Code enhancements stay optional and isolated. Where a Claude Code-only mechanism (like hook-based enforcement in hooks-enforcement) can mechanically assist a rule, it lives under skills/models/claude/ as an opt-in add-on — the underlying contract in coding-standards/ai-standards.md works the same with or without it, on any tool.

Recent additions

• Tool-routed installs. install-kit now detects the target tool and routes both the skill root and the session-protocol file (CLAUDE.md for Claude Code, AGENTS.md for OpenCode/Codex) — a contract in a file the tool never opens can no longer pass as a successful install. Parallel-session claims moved from a shared locks.md table to atomic handover/locks.d/ claims with per-session ids, so concurrent Claude Code and OpenCode sessions can't silently clobber each other.
• Agent-tool discipline. agent-usage defaults every session to inline work — no Agent-tool delegation — unless the user names an agent explicitly or a scope-anchored need is judged and approved first; mode-kernel is the shared table governing how that gate (and three other skills' own stop-and-wait gates) behaves across autonomous, planning, and background session modes.
• Skill ablation. skill-ablation is a periodic companion to memory-gardener: archive the accumulated session-protocol file/skills/hooks, run real work with none of it, and restore only what repeated real-world evidence proves is still needed — catching obsolete instructions a line-count cap alone can't.
• Live-pulled Shopify skills. shopify-toolkit-install clones Shopify's own AI toolkit straight into a target project's skill root at install time — this kit never stores or forks Shopify-authored files, so authorship and their telemetry hook stay exactly where they belong.
• Tool-compatibility checkpoint. Any skill that depends on a mechanism only one AI coding tool provides (Claude Code hooks, for example) now declares it structurally via a compat: <tool>-only frontmatter field, checked automatically by install-kit at install time — so tool-specific features get flagged and kept out of incompatible projects instead of failing silently.
• Opt-in addons. system1-prefilter introduces a fourth skill-scope category — a capability that's never stack-detected or installed by default, only turned on when the user explicitly asks, then stays pinned as their own durable choice. skill-scope now tracks this category alongside permanent-pinned and per-session-archivable.

Repository layout

skills/
  README.md        ← catalog + install instructions (start here)
  models/          ← per-model protocols + fleet routing (claude/ lineup + Claude Code hooks, opencode/ open-weight)
  workflow/        ← intent capture, planning, handover, debugging, budget, commits, pre-merge checks, perf, agent-usage discipline, session-mode gating, parallel sessions, e2e scaffolding
  standards/       ← the coding-standards enforcement skill
  memory/          ← repo-committed memory, knowledge gardening, periodic skill ablation
  stack/           ← Three.js, Astro, Shopify (live-pulled from Shopify's own toolkit, never forked)
  libraries/       ← skills for the author's own libraries (strata-css, triforge)
  meta/            ← skill-writer (quality bar), install-kit (installer), skill-scope (pin/archive/addon classification)
  addons/          ← opt-in, user-opted capabilities, never installed by default (system1-prefilter)
migrations/
  RENAMES.md       ← skill rename ledger — install-kit reads it to migrate old installs (never delete)
coding-standards/
  index.md         ← system map: layers, reading order, file-to-role mapping (start here)
  *-standards.md   ← global rules per discipline (css, html, js/ts, git, seo, a11y, qa, ai, …)
  */               ← file-role partials per discipline
  frameworks/      ← framework additions/overrides (astro, bootstrap)
  tooling/         ← lint configs enforcing the machine-checkable rules
  CLAUDE.example.md← session-protocol template — wire into CLAUDE.md (Claude Code) or AGENTS.md (OpenCode)

Maintaining the kit

• This repo is the canonical home. Improve skills here, then re-run the install into projects; the installer diffs before overwriting so locally-evolved copies are never silently clobbered.
• New skills follow the quality bar in skills/meta/skill-writer.md.
• Standards edits follow the repo's own principles (index.md → Principles): testable, declarative, wrong/right examples mandatory. Machine-checkable rule changes update coding-standards/tooling/ in the same commit.
• learnings.md files are gitignored — they belong to the project that earned them. Lessons worth keeping forever get promoted into skill bodies (see memory-gardener).

License

MIT. The license's "software" wording legally covers this collection of markdown skills, standards, and configs — use, adapt, and redistribute freely with attribution.