pastiche CLAUDE.md

pastiche CLAUDE.md is an instructions file for coding agents from retz8/pastiche. It costs 761 tokens per session, scanned A, original, MIT.

A design-system guide for AI coding agents building frontend interfaces with an existing component library and visual language. It uses catalogs and rules to show which components and design tokens fit each situation.

In plain words
What is it for?
Implementing frontend screens within an existing design system, choosing approved components and tokens, and reviewing code for faithful design-system use.
Why use it?
It keeps generated UI consistent with the product's established design instead of letting the agent invent unrelated styles or components.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/retz8/pastiche/claude-md
Clone the repo
git clone --depth 1 https://github.com/retz8/pastiche

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for pastiche CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/retz8/pastiche/claude-md.svg)](https://agentmods.dev/instructions/retz8/pastiche/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/retz8/pastiche/claude-md"><img src="https://agentmods.dev/badge/instructions/retz8/pastiche/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 761 This file is loaded in full into every session.
When invoked 761 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00761 $0.00761
Opus 5 $0.00380 $0.00380
Sonnet 5 $0.00152 $0.00152
Haiku 4.5 $0.00076 $0.00076

Measured 4d ago against content hash b251f1b91f79, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

pastiche CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 38 lines

How it starts

The opening of the file, as written. The whole thing — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.

pastiche

What this project is

Pastiche is a Claude skill (and forthcoming OSS package) for faithful design-system execution by LLM coding agents. Given a frontend task, it produces code that follows an established design system + component library — not by inventing, but by faithfully executing within the established vocabulary.

The skill operates over three documents that form an epistemological hierarchy:

  • FACT.md — mechanical catalog of every token, component, and prop. Auto-extracted from the codebase.
  • KNOWLEDGE.md — curated scenario→atom mappings (the "absent designer") + brand-identity prose.
  • WISDOM.md — atom-tagged rules. Atom-intrinsic only; scenario-conditional rules live in KNOWLEDGE.

The runtime mechanism is a bounded doubt-defense loop: pastiche-implementer-round1 produces code → pastiche-reviewer raises strict-YAML doubts → pastiche-implementer-round2 corrects or defends. Depth comes from the dialogue, not from a heavy reviewer.

Required reading: spec.md (philosophy), _dev/TODO.md (post-v1 backlog).

How to work in this repo

Daily work harness

Full design: _dev/docs/harness.md. Summary:

  • Hierarchy: _dev/TODO.md groups (one minor version each) → tasks (parent GH issue + feature branch <version>/<task-name>) → sub-tasks (child GH issues + <feat>/<sub-task-name> branches).
  • Task kickoff: pick-up-task (marks [WIP], handles branch) → grill-megrill-to-spec → plan if needed (plan docs at _dev/docs/<branch-name>/; small patches at _dev/docs/patches/) → divide-task (sub-task issues labeled autonomous-ready / decision-needed / blocked-by-<issue#>, mirrored into TODO.md).
  • Execution: nightly routine picks only autonomous-ready, unblocked sub-tasks of the [WIP] task; PRs sub-task → feat and auto-merges. Live sessions use worktrees off the feat branch unless told otherwise. When a blocker completes, remove its blocked-by-# label from blocked issues.
  • Review: Monday session reviews the accumulated feat diff. Small fixes directly on feat; substantial rework → new sub-task issue. Nothing reaches main without human review of the feat→main PR.
  • Commits: conventional commits — <type>(<task-name>): on task work, <type>: otherwise.
  • Releases: tasks don't bump versions. Group complete → release-pastiche on main (bump → build → CHANGELOG → tag → GH Release → close milestone). Patch = hotfix only; in-flight feat branches absorb hotfixes via rebase-feat-with-main.

Read the full file on GitHub · 38 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 38 lines · 761 tokens per session scan A b251f1b91f79

Subscribe to this mod's changes

pastiche CLAUDE.md is an instructions file published in the GitHub repository retz8/pastiche (4 stars, last pushed 2mo ago), licensed MIT. It adds 761 tokens to every session, about $0.0038 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other instructions, from other repositories