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 instructions/let-sunny/canicode/claude-mdgit clone --depth 1 https://github.com/let-sunny/canicodeWrote 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.
[](https://agentmods.dev/instructions/let-sunny/canicode/claude-md)<a href="https://agentmods.dev/instructions/let-sunny/canicode/claude-md"><img src="https://agentmods.dev/badge/instructions/let-sunny/canicode/claude-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.02390 | $0.02390 |
| Opus 5 | $0.01195 | $0.01195 |
| Sonnet 5 | $0.00478 | $0.00478 |
| Haiku 4.5 | $0.00239 | $0.00239 |
Grade A, and why
canicode 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 5d 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 — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CanICode
A CLI tool that analyzes Figma design structures to provide development-friendliness and AI-friendliness scores and reports.
Core Goal
Make the Figma file information-complete so figma-implement-design produces accurate code with fewer gotchas.
canicode's role is upstream of code generation: diagnose where design information is missing (analyze), elicit the missing answers from the user (gotcha-survey), and write those answers back into the Figma design (canicode-roundtrip). Once the design re-analyzes clean, the downstream code-generation step runs in Figma's official figma-implement-design skill — canicode does not own that step (see ADR-013 for the scope boundary).
The design-tree format used internally by analysis is a curated, CSS-ready representation; ablation experiments use it as a controlled measurement input. The framing information curation > information abundance still drives rule design — fewer information gaps in the source design means cleaner downstream code.
See Experiment Wiki for detailed data and methodology.
Target Environment
The primary target is teams with designers where developers (+AI) implement large Figma pages:
- Page scale: 300+ nodes, full screens, not small component sections
- Component-heavy: Design systems with reusable components, variants, tokens
- AI context budget: Large pages must fit in AI context windows — componentization reduces token count via deduplication
- Not the target: Individual developers generating simple UI with AI — they don't need Figma analysis
This means:
- Component-related rule scores (missing-component, etc.) should NOT be lowered based on small fixture calibration
- Token consumption is a first-class metric — designs that waste tokens on repeated structures are penalized
- Calibration fixtures must be large-scale (270+ nodes) — experiments showed small fixtures (50-100 nodes) produce misleading results
no-auto-layoutis the single highest-impact rule (score -10) — empirically validated via ablation experiments
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.
- 5d ago First seen · 202 lines · 2,390 tokens per session scan A d630610d2ab1
canicode CLAUDE.md is an instructions file published in the GitHub repository let-sunny/canicode (10 stars, last pushed 2mo ago), licensed MIT. It adds 2,390 tokens to every session, about $0.0120 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 instructions, from other repositories
Figma-Context-MCP CLAUDE.md
Instructions for GLips/Figma-Context-MCP, covering framelink mcp for figma, build & development commands, running the server, running a single test and releasing.
figwright AGENTS.md
AGENTS.md instructions for awdr74100/figwright, covering agents.md, architecture, layout, tech stack and commands.
figwright CLAUDE.md
Claude Code instructions for awdr74100/figwright: Read AGENTS.md first — it's the single source of guidance for this repo: architecture, layout, tech stack, commands, conventions, release flow, and gotchas.
bridge CLAUDE.md
Instructions for noemuch/bridge, covering bridge ds — claude code instructions, mission, the 3 pillars, design principle: subtraction by default and architecture.
SoDam-Design-Kit AGENTS.md
AGENTS.md instructions for sodam-ai/SoDam-Design-Kit, covering sodam-design-kit — codex 규칙, 목표, 작업 순서 (고정), 절대 금지 and 완료 조건.
aiconnect-figma-mcp CLAUDE.md
Instructions for guptaprakhariitr/aiconnect-figma-mcp, covering claude.md, project overview, build & development commands, architecture and mcp server (src/aiconnectmcp/server.ts).