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.
npx agentmods add skills/evalstate/birch-html/skillnpx skills add evalstate/birch-html --skill skillgit clone --depth 1 https://github.com/evalstate/birch-htmlWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00137 | $0.02816 |
| Opus 5 | $0.00068 | $0.01408 |
| Sonnet 5 | $0.00027 | $0.00563 |
| Haiku 4.5 | $0.00014 | $0.00282 |
Grade A, and why
birch-html 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 2d 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.
How it starts
The opening of the file, as written. The whole thing — 231 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Birch HTML
Create one complete, locally opening Birch HTML artifact. Optimize for complete HTML, preserved Birch CSS placeholder, source-grounded claims, 390px mobile safety, strict component contracts, and minimal page-local CSS.
Output contract
- Return a full HTML document only: no prose, Markdown fences, or partial HTML.
- If writing to disk, validate and return only the artifact path.
- Keep the placeholder exactly until postprocessing:
<style data-birch-system>__BIRCH_SYSTEM_CSS__</style>. - Do not paste bundled Birch CSS or modify canonical CSS/scripts/resources.
- No network assets: no remote URLs, CDNs, fonts, iframes, images, scripts, or styles.
- Use
<main class="page stack" data-gap="lg">as the outer visible shell. - Do not replace the Birch shell with a custom app/dashboard shell. Do not make
body,.container,.report, or.dashboardthe visible page shell. - Do not invent Birch classes/CSS variables or use Birchline-only classes. In
particular, avoid custom path/helper classes such as
.file-pathand avoid variables such as--color-*,--size-*,--s-*,--bg-alt, or--fg-mutedunless they are already defined by Birch. - Use flat token surfaces only: no gradients, glassmorphism, glow backgrounds, decorative background images, or custom dashboard skins.
- Birch style forbids gradients for card, panel, page, and callout backgrounds
and for body/page backgrounds (for example
linear-gradient,radial-gradient, orconic-gradient). Use flat Birch token surfaces instead. - For diffs, never paste raw unified-diff text into
<pre>or code blocks. Represent patch lines with.diff-row; the only literal+/-marker should be inside<span class="mark">.
Workflow
- Read sources first: files, diffs, data, logs, commands, screenshots, session notes, and local links before writing claims.
- Make a terse brief: audience, purpose, exact paths/values, caveats, unknowns, sections, and chart intent.
- Ask only if missing audience, scope, sources, or output path would materially change the result; otherwise proceed and label assumptions.
- Load only needed recipes:
numeric-data,code-review,pr-change-writeup,implementation-plan,module-explainer,process-explainer,design-directions,status-incident-report,flow-diagram,slide-deck,benchmark-comparison, ordesign-system-reference. - If no path is specified and filesystem access exists, write a descriptive
kebab-case
.htmlfile in the current directory. - Prefer copying
skill/resources/template.html; replace body content only. Then runuv run skill/scripts/finish_birch_html.py <output.html>. - Validate before returning: doctype, html/head/body, viewport, title,
.page, Birch CSS placeholder/embedded CSS, closing</html>, no network assets, no unknown classes/variables, strict children, wrapped long content, no 390px overflow, and no raw patch marker lines. If available run:uv run --with pillow python skill/scripts/check_birch_renderings.py --artifact <output.html>.
What ships with it
19 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- assets/birch-system.css 31 KB
- recipes/benchmark-comparison.md 5.2 KB
- recipes/code-review.md 4.6 KB
- recipes/design-directions.md 1.4 KB
- recipes/design-system-reference.md 6.6 KB
- recipes/flow-diagram.md 1.8 KB
- recipes/implementation-plan.md 5.9 KB
- recipes/module-explainer.md 3.2 KB
- recipes/numeric-data.md 3.3 KB
- recipes/pr-change-writeup.md 1.5 KB
- recipes/process-explainer.md 1.3 KB
- recipes/slide-deck.md 1.4 KB
- recipes/status-incident-report.md 1.6 KB
- resources/template.html 972 B
- scripts/birch_mpl.py 3.4 KB runs code
- scripts/birch-copy.js 1.9 KB runs code
- scripts/check_birch_renderings.py 75 KB runs code
- scripts/finish_birch_html.py 2.4 KB runs code
- styles/birch-system.css 31 KB
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.
- 2d ago First seen · 231 lines · 137 tokens per session scan A f398706fa218
birch-html is a skill published in the GitHub repository evalstate/birch-html (5 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 137 tokens to every session and 2,816 once invoked, about $0.0007 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.
Other skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
babysit-pr
Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…
imagegen
Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…
next-cache-components-optimizer
Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…