Back to skill

Security audit

Urufu Agent

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent and disclosed, but it handles wallet signing authority and paid mints in ways that need careful review before use.

Review before installing. Use only a dedicated low-fund wallet or enforceably limited session key, do not put private keys in shell startup files, avoid mint-paid until the cooldown bug is fixed, inspect every transaction/signature before approval, and update the vulnerable dependency lock before relying on long-running network use.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/mint-paid.mjs:123
Finding

Paid mint transactions bypass the mandatory 24-hour cooldown

Content
View full analysis
p.phaseId === preview.phaseId); if (!target) return { skipped: true, reasons: ['phase not found'], preview }; const allowlist = await fetchMintAllowlistProof(config.siteUrl, target.info.merkleRoot, config.wallet); if (!allowlist) return { skipped: true, reasons: ['not on allowlist'], preview }; const maxPayment = await resolveMaxPayment(client, config, target, quantity); const [nonce, paramsHash] = await Promise.all([ client.readContract({ address: config.minter, abi: minterAbi, functionName: 'mintNonce', args: [getAddress(config.wallet)], }), client.readContract({ address: config.minter, abi: minterAbi, functionName: 'mintParamsHash', args: [target.phaseId, quantity, maxPayment], }), ]); const { tickets, signatures } = await requestRandomnessTickets(config, { nonce, paramsHash, phaseId: target.phaseId, phase: target.info, quantity, maxPayment, }); const walletClient = createWalletClient({ account, ch ...[truncated 3286 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
references/onboarding.md:78
Finding

Onboarding recommends shell-wide exposure of a spend-capable private key

Content
View full analysis
for claim/mint i need a **signing key on the agent vm** — use a session key or hot wallet, **not** ur main vault seed. | ✅ put key here | ❌ never here | |---|---| | agent VM env: `export URUFU_PRIVATE_KEY=0x…` in `~/.bashrc` / `~/.zshrc` | this chat | | agent VM file: `~/.urufu-agent/env` (chmod 600) | git repo | | OpenClaw/Hermes secrets / env in skill host config | screenshot to discord | | | frontend `.env.local` (thats the website, not ur bot) | Suggested file `~/.urufu-agent/env`: ```bash export URUFU_WALLET=0xYourAddress export URUFU_PRIVATE_KEY=0xYourSessionKey export URUFU_RPC_URL=https://base-mainnet.g.alchemy.com/v2/YOUR_KEY export URUFU_INDEXER=https://neochibi-api.radbro.xyz export URUFU_SITE_URL=https://www.urufu.xyz ``` Then: `source ~/.urufu-agent/env` before steward commands, or add `source` line to shell rc. ``` ### Technical Analysis The onboarding documentation recommends placing `URUFU_PRIVATE_KEY` directly in `.bashrc` or `.zshrc`, or automatically sourcing a plaintext key file from shell initialization. A shell startup file is loaded for many interactive shell sessions. Once exported, the private key is inherited by every descendant process unless the environment is explicitly sanitized. This expands access beyond the steward process and violates least-privilege credential handling. Although the guidance correctly warns against using a main seed phrase and recommends mode `0600`, the configured key can still authorize direct wallet operations. The implementation uses it to create a normal `viem` account and wallet client, including paid ETH and URU transactions. The software does not technically constrain the key to claim-only EIP-712 intents. File permissi ...[truncated 1596 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (36)

Known Vulnerable Dependency: ws==8.20.1 — 1 advisory(ies): CVE-2026-48779 (ws: Memory exhaustion DoS from tiny fragments and data chunks)

High
Category
Supply Chain
Confidence
92% confidence
Finding

The lockfile pins ws to version 8.20.1, which the provided advisory identifies as vulnerable to memory exhaustion denial of service via tiny fragmented frames/data chunks. This package is pulled in through viem and is relevant because the skill interacts with blockchain/RPC infrastructure, where WebSocket connections are commonly used for subscriptions and event streams, making a reachable DoS condition more plausible.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
95% confidence
Finding

The documentation instructs users to persist a private key in a plaintext shell env file under the home directory. Even with chmod 600, storing signing keys in a long-lived file increases exposure to local compromise, malware, backup leakage, shell sourcing mistakes, and accidental inclusion in logs or tooling, which is especially dangerous here because the skill performs real blockchain signing operations.

Content

Scanner excerpt · README.md (reported line 56)May include surrounding context.

3. Configure env (your agent VM — not in chat)

Create ~/.urufu-agent/env (chmod 600):

bash
export URUFU_WALLET=0xYourAddress

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The trigger list includes broad phrases such as 'claim URU' and 'agent play urufu' that could plausibly occur in normal conversation and unintentionally activate a skill capable of wallet-related actions. In this context, accidental invocation is more dangerous than usual because the skill can monitor holdings and, with further prompting or existing configuration, prepare or execute blockchain actions affecting user assets or relayer resources.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file states that the agent "speaks in urufu voice for all game talk," which imposes a fixed language/style policy rather than offering a user choice. This is a natural-language policy issue because the skill requires a specific communication style by default without explicit opt-in.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README instructs users to export URUFU_PRIVATE_KEY directly in the shell without an explicit warning about credential sensitivity, secure storage, or the risks of environment-variable exposure. In a wallet/agent skill that performs on-chain claims and mints, normalizing raw private-key handling materially increases the chance of key compromise through shell history, logs, process inspection, or unsafe operator practices.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The description states broad read and write surfaces for AI agents without defining narrow triggers, approval boundaries, or disallowed operations. In agent environments, vague capability framing increases the chance of overbroad autonomous actions such as submitting wallet-related requests or invoking unrelated endpoints when only a limited user task was intended.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · docs/api/openapi.yaml (reported line 15)May include surrounding context.

yaml
**Agent safety**
    - Claim at most once every 6 hours per wallet (client-side timer required).
    - Never store relayer keys; sign ClaimIntent with user wallet only.
    - See `.claude/skills/urufu-agent-play/SKILL.md` and `scripts/agent/`.

  contact:
    name: urufu gēmu

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/bankr.md (reported line 61)May include surrounding context.

md
**Agent safety**
    - Claim at most once every 6 hours per wallet (client-side timer required).
    - Never store relayer keys; sign ClaimIntent with user wallet only.
    - See `.claude/skills/urufu-agent-play/SKILL.md` and `scripts/agent/`.

  contact:
    name: urufu gēmu

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The write endpoints accept signed wallet intents for gasless claim and mint operations, but the spec does not pair them with a strong user-facing warning that these calls can trigger on-chain state changes and consume relayer resources. In a wallet-connected agent context, this omission is especially risky because users may not realize that generating and submitting signatures authorizes blockchain actions, even if gas is paid by a relayer.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The OpenAPI spec exposes multiplayer room creation, join, snapshot, and documented WebSocket session capabilities that are outside the skill metadata's stated portfolio/claim/mint scope. For an agent-integrated skill, this scope expansion is dangerous because it enables unadvertised network actions, session token handling, and interactive behaviors that users or orchestrators may not expect or consent to.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest description contains several broad trigger phrases such as "claim URU," "mint chibi," and "agent play urufu" that can plausibly match ordinary user conversation and cause the skill to activate unexpectedly. Because this skill is tied to wallet operations, unintended invocation increases the chance of prompting sensitive actions or initiating checks and transaction-related flows in a higher-risk financial context.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · README.md (reported line 56)May include surrounding context.

Put secrets on the agent machine, not in chat:

bash
# ~/.urufu-agent/env  (chmod 600)
export URUFU_WALLET=0xYourAddress
export URUFU_PRIVATE_KEY=0xYourSessionKey
export URUFU_RPC_URL=https://base-mainnet.g.alchemy.com/v2/YOUR_KEY

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · references/BOOTSTRAP.md (reported line 28)May include surrounding context.

Put secrets on the agent machine, not in chat:

bash
# ~/.urufu-agent/env  (chmod 600)
export URUFU_WALLET=0xYourAddress
export URUFU_PRIVATE_KEY=0xYourSessionKey
export URUFU_RPC_URL=https://base-mainnet.g.alchemy.com/v2/YOUR_KEY

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/BOOTSTRAP.md (reported line 34)May include surrounding context.

export URUFU_RPC_URL=https://base-mainnet.g.alchemy.com/v2/YOUR_KEY

text

Get RPC: [alchemy.com](https://www.alchemy.com) → Create app → **Base Mainnet** → copy HTTPS URL.  
See `references/onboarding.md` for step-by-step + Bankr path.

---

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · references/BOOTSTRAP.md (reported line 59)May include surrounding context.

md
1. ask: clawhub, hermes, or bankr?
2. ask wallet address (0x only — NEVER seed in chat)
3. explain how to get Base RPC (Alchemy: create app → Base Mainnet → HTTPS url)
4. explain where to put session key: ~/.urufu-agent/env on agent VM, chmod 600 — NOT in chat, NOT in git
5. run check, report herd + yield + next claim time in voice

Wallet: YOUR_WALLET_ADDRESS (or ask me)

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · references/onboarding.md (reported line 83)May include surrounding context.

md
1. ask: clawhub, hermes, or bankr?
2. ask wallet address (0x only — NEVER seed in chat)
3. explain how to get Base RPC (Alchemy: create app → Base Mainnet → HTTPS url)
4. explain where to put session key: ~/.urufu-agent/env on agent VM, chmod 600 — NOT in chat, NOT in git
5. run check, report herd + yield + next claim time in voice

Wallet: YOUR_WALLET_ADDRESS (or ask me)

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · references/onboarding.md (reported line 156)May include surrounding context.

md
1. ask: clawhub, hermes, or bankr?
2. ask wallet address (0x only — NEVER seed in chat)
3. explain how to get Base RPC (Alchemy: create app → Base Mainnet → HTTPS url)
4. explain where to put session key: ~/.urufu-agent/env on agent VM, chmod 600 — NOT in chat, NOT in git
5. run check, report herd + yield + next claim time in voice

Wallet: YOUR_WALLET_ADDRESS (or ask me)

External Transmission

Medium
Category
Data Exfiltration
Confidence
84% confidence
Finding

The documentation instructs users to store a high-privilege Bankr API key in an environment variable and use it in direct requests to a third-party service. If downstream skills, logs, shell history, prompt injection, or command echoing expose that variable, an attacker could use the key to access the managed wallet API and authorize sensitive wallet actions.

Content

Scanner excerpt · references/bankr.md (reported line 33)May include surrounding context.

bash
export BANKR_API_KEY=bk_...
# Wallet address from: curl https://api.bankr.bot/agent/user -H "X-API-Key: $BANKR_API_KEY"
export URUFU_WALLET=0x...   # Bankr Base address

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

This flow explicitly sends typed signing payloads to Bankr and then transmits the resulting signature to urufu relay endpoints, enabling real wallet-affecting actions through external services. In an agent setting, any prompt injection, compromised dependency, or insufficient user-confirmation logic could trick the agent into signing unintended messages or relaying signatures for unauthorized claims/mints.

Content

Scanner excerpt · references/bankr.md (reported line 52)May include surrounding context.

md
1. Read yield via indexer + on-chain preview (optional RPC or steward check)
2. Build ClaimIntent typed data (see urufu skill)
3. `POST https://api.bankr.bot/agent/sign` with typed data
4. `POST https://www.urufu.xyz/api/gasless/claim` with signature

Same timers apply: **6h claim**, **24h mint**, explicit user commands only.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The document states the voice is mandatory for all urufu agent talk and requires a specific register whenever discussing urufu-related topics. This forces a language/style choice by default rather than offering the user a choice or opt-in, which fits the locale/language policy concern.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/onboarding.md (reported line 112)May include surrounding context.

bash
export BANKR_API_KEY=bk_...
export URUFU_WALLET=0x...   # from GET https://api.bankr.bot/agent/user

Indexer reads work without RPC. Signing goes through Bankr /agent/sign.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The example trigger phrase "help me play urufu with you" is broad enough that ordinary conversation can activate the skill unexpectedly. Because this skill can later guide wallet setup and prepare claim/mint flows, accidental activation increases the chance of users being prompted for wallet/RPC details in contexts where they did not explicitly intend to use the urufu agent.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
74% confidence
Finding

The onboarding flow encourages persistent storage of RPC and signing credentials on the agent VM and a conversational handoff of setup state ("say done/ready"), which can normalize long-lived secret retention for a wallet-capable agent. In a skill that can claim, mint, and sign on-chain actions, persistent session material increases the blast radius if the host, logs, or agent environment are later compromised.

Content

Scanner excerpt · references/onboarding.md (reported line 152)May include surrounding context.

md
**Human:** 0xabc…

**Agent:** got it. next: base rpc. urufu dont host one — grab free alchemy base mainnet (alchemy.com → create app → base → copy https url). paste it here OR set `URUFU_RPC_URL` on ur vm and say "done"

**Human:** done, set on my vm

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The README instructs users to set URUFU_PRIVATE_KEY for claim/mint operations but does not clearly warn that this is a highly sensitive credential that can authorize on-chain actions and potentially expose wallet funds if mishandled. In an agent-oriented skill, users may copy this into shell history, logs, config files, or broader-permission environments, which increases the chance of credential leakage or misuse.

Content

No source excerpt is available for this finding.