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 skills add 0xIntuition/agent-skills --skill erc8004-agent-layergit clone --depth 1 https://github.com/0xIntuition/agent-skillsWrote 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/skills/0xintuition/agent-skills/erc8004-agent-layer)<a href="https://agentmods.dev/skills/0xintuition/agent-skills/erc8004-agent-layer"><img src="https://agentmods.dev/badge/skills/0xintuition/agent-skills/erc8004-agent-layer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/0xintuition/agent-skills/erc8004-agent-layer"><img src="https://agentmods.dev/badge/skills/0xintuition/agent-skills/erc8004-agent-layer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00115 | $0.02742 |
| Opus 5 | $0.00057 | $0.01371 |
| Sonnet 5 | $0.00023 | $0.00548 |
| Haiku 4.5 | $0.00012 | $0.00274 |
Grade A, and why
erc8004-agent-layer 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 11d 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 — 293 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ERC-8004 Agent Layer
Produce one deterministic semantic plan for an ERC-8004 partner integration. Treat this skill as a narrow planner, not an execution agent.
This skill is self-contained for semantic planning. The core intuition skill
is not a prerequisite for planning and enters only in a separately authorized
later protocol phase. The canonical human guide is
https://docs.intuition.systems/docs/erc-8004-agent-layer.
Phase boundary
The initial partner request is always the semantic-planning phase, even when
the user says “prepare everything,” “take care of the integration,” “complete
it,” or similar broad language.
In this phase:
- Run the canonical planner command.
- Save its unmodified JSON to one canonical result file and return the path plus a concise explanation.
- Stop the turn.
- Do not load or invoke the core
intuitionskill. - Do not call
pinThingor any other mutation. - Do not query write costs, preview writes, resolve or mint atoms, calculate write calldata, prepare unsigned transactions, inspect wallet secrets, sign, or broadcast.
Protocol preparation is a separate later phase. It requires both:
- A canonical plan produced by this skill with
status: semantic_plan_readyand a successful fresh validation. - A new user message, after the plan was returned, explicitly asking to continue that plan into protocol preparation.
A broad request in the initial message is not this later authorization.
Approval ladder
Treat the requested network as plan data, never as write authorization. Classify every ERC-8004 request into the lowest applicable stage:
semantic-planning: run GraphQL reads and return the canonical plan. This is the only stage allowed in the initial user turn.protocol-preview: after a new request naming the returned plan, revalidate it, read costs, prepare unsigned transactions, and simulate. Do not pin, sign, or broadcast.metadata-pinning: treatpinThingas a persistent external mutation. First return a pinning approval request that identifies the plan integrity SHA and the exact ordered Thing payloads. Only a later message explicitly approving that presented request authorizes pinning. Pinning grants no chain-write authority.testnet-execution: require a separate, one-shot approval bound to the exact simulated testnet transaction set on chain13579.mainnet-preview: require a fresh mainnet semantic plan and a new request. Testnet plans, simulations, and approvals do not carry forward.mainnet-execution: require a final, one-shot approval bound to the exact simulated mainnet transaction set on chain1155.
What ships with it
10 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.
- agents/openai.yaml 233 B
- README.md 5.8 KB
- reference/graphql.md 2.7 KB
- reference/partner-manifest.example.json 1.0 KB
- reference/registry.json 21 KB
- reference/workflow.md 13 KB
- scripts/build-partner-plan.mjs 3.3 KB runs code
- scripts/partner-plan.mjs 59 KB runs code
- scripts/validate-partner-plan.mjs 1.4 KB runs code
- scripts/verify-registry.mjs 3.9 KB runs code
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.
- 11d ago First seen · 293 lines · 115 tokens per session scan A be8f6c84a9c4
erc8004-agent-layer is a skill published in the GitHub repository 0xIntuition/agent-skills (22 stars, last pushed 1mo ago), licensed MIT. It adds 115 tokens to every session and 2,742 once invoked, about $0.0006 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 skills, from other repositories
signature-replay
Signature replay attacks — missing nonces, missing chain ID, ecrecover zero address, signature malleability, cross-chain replay.
official-sui-skills
Pointer to the official Mysten Labs skills for building on Sui — language fundamentals, object model, PTBs, SDKs, publishing, upgrades, frontend integration, accessing on-chain data. Maintained upstream at github.com/MystenLabs/skills; pinned to the same ref the audit catalog derives from (see…
bridge
Cross-chain token transfers using Wormhole and CCTP.
onchainkit
Build onchain apps with Coinbase's OnchainKit React components - wallets, swaps, NFTs, payments.
marginfi
MarginFi — Solana lending and borrowing.
crypto-market-rank
Crypto market rankings and leaderboards. Query trending tokens, top searched tokens, Binance Alpha tokens, tokenized stocks, social hype sentiment ranks, smart money inflow token rankings, top meme token rankings from Pulse launchpad, and top trader PnL leaderboards. Use this skill when users ask about token rankings…