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/glebis/humane-agentic-design/claude-mdgit clone --depth 1 https://github.com/glebis/humane-agentic-designWrote 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/glebis/humane-agentic-design/claude-md)<a href="https://agentmods.dev/instructions/glebis/humane-agentic-design/claude-md"><img src="https://agentmods.dev/badge/instructions/glebis/humane-agentic-design/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 | $0.02996 | $0.02996 |
| Opus 5 | $0.01498 | $0.01498 |
| Sonnet 5 | $0.00599 | $0.00599 |
| Haiku 4.5 | $0.00300 | $0.00300 |
Grade A, and why
humane-agentic-design 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 yesterday.
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 — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Guidance for agents working in this repository.
What this repository is
humane — a design method cycle packaged as a Claude Code plugin and its own
single-entry marketplace, portable to other agents. It covers the work around an
interface: what to build (jtbd), for whom (persona-review,
respondent-panel), under which system (design-tokens) and constraints
(layout-rules), in which words (ux-writing), audited (nielsen-heuristics)
and proven (before-after). Ring 2 (brandkit, brand-illustrate) handles
visual identity. design-frameworks is a separate adapter layer: it fits an
already-decided implementation to an existing design system without becoming a
method step or a review domain.
Skills live in humane/skills/<name>/SKILL.md; the bundled agent is
humane/agents/synthetic-respondent.md. Bump version in
humane/.claude-plugin/plugin.json when plugin users should receive an update,
and add the entry to CHANGELOG.md in the same commit — its "How to write an
entry" section owns the format (capability heading, numbered steps, ### Fixed
lists, and what never gets documented).
Rule ownership
Each rule lives in exactly one skill. Other skills reference it by skill name in backticks and name only the handoff — never restate or override it.
| Skill | Owns |
|---|---|
using-humane |
Routing only: which skill owns a task, and what the declared graph says comes next. Owns the graph keys (handoffs/accepts/orchestrates) and their reader, scripts/graph.py. Owns no method rules, and no edge — each skill declares its own. |
setup |
Environment and configuration: the settings table, its resolution order, dependency checks, and the install commands for each gap. Owns no method rules. |
jtbd |
The corpus: jobs, switch forces, ODI outcomes, evidence ledger, granularity and jargon gates. Every downstream claim cites it. |
persona-review |
Expert stakeholder critique of a document that is meant to be studied |
respondent-panel |
Gut reactions from strangers to a user-facing artifact, in isolated contexts. Never rewrites. |
prototype |
Pre-system fidelity: the ASCII → SVG click-dummy → HTML ladder, the one-question rule, the self-containment contract, and the "what is fake" disclosure. Owns no token, no final wording, and never reviews its own output. |
design-frameworks |
Implementation fit: preset selection, native capability discovery, component/token/template mapping, guarded framework writes, and framework compliance. Owns no job, evidence, concept, prototype, token authorship, or product-review rule. |
design-tokens |
The token set: DTCG structure, resolution, CSS/DESIGN.md export, the brand block, and all color measurement and remediation (tokens contrast) |
type-specimen |
Typeface selection: the specimen page, its ten-slot copy contract, script and glyph coverage checks, and what a specimen may and may not conclude. Owns no token. |
layout-rules |
Structural and interaction defect classes for tool/dashboard UIs, and the de-slop avoid-list |
ux-writing |
The source wording of every user-facing string, and what each string must accomplish — including documentation prose (README, docs, release notes) |
nielsen-heuristics |
Formal usability inspection against the 10 heuristics, with severity and evidence locators |
walkthrough |
Task completion — whether a specific person with a specific job can get through a specific interface. Also owns the driven-mode procedure (references/driven.md): browser tool ladder, device matrix, screenshot evidence contract, mode gate |
review |
Orchestration only: scope, mode, domain order, consolidation, coverage honesty, and the single final verdict. Owns no domain rules. |
before-after |
The felt transformation claim, and the proof that a change worked |
brandkit |
Identity exploration — competing directions for a brand that does not exist yet |
brand-illustrate |
Asset production under an existing token contract |
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.
- yesterday Changed · +11 lines · +187 tokens per session e375f35aaf2d
- 6d ago First seen · 179 lines · 2,809 tokens per session scan A 3fb2680f8097
humane-agentic-design CLAUDE.md is an instructions file published in the GitHub repository glebis/humane-agentic-design (27 stars, last pushed 4d ago), licensed MIT. It adds 2,996 tokens to every session, about $0.0150 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 2026-08-30.
Other instructions, from other repositories
figma-console-mcp CLAUDE.md
Claude Code instructions for southleft/figma-console-mcp, covering figma console mcp, build & test, release process, known issues and architecture.
starwind-ui AGENTS.md
AGENTS.md instructions for starwind-ui/starwind-ui, covering agents.md, project shape, useful commands and contribution conventions.
ACCESSIBILITY.md AGENTS.md
AGENTS.md instructions for mgifford/ACCESSIBILITY.md, covering repository instructions for ai coding agents, scope, when each instruction file is useful, project mission and status and required project context.
Codex-Skills AGENTS.md
AGENTS.md instructions for TheGoat395/Codex-Skills, covering repository operating notes, required checks, evidence rules and change hygiene.
SidewalkWebpage css.instructions.md
Instructions for ProjectSidewalk/SidewalkWebpage: --text- font shorthands (font: var(--text-body-regular);) — not raw --font- plus hand-picked size/weight. Flag hardcoded hex colors or hand-assembled font stacks where a token fits.
OpenGUI CLAUDE.md
Claude Code instructions for Core-Mate/OpenGUI, covering opengui release - repository guide for claude code / codex, 1. repository positioning, 2. default working rules, 3. high-level architecture and 4. directory map.