Back to skill

Security audit

Agent Treasury

Security checks for vulnerabilities and agentic risk

Overview

This crypto-wallet skill is mostly coherent, but it includes high-risk mainnet transfer instructions and hidden external agent-registration metadata that deserve user review before installation.

Install only if you are prepared to treat it as high-risk wallet guidance: do not paste private keys into source files, do not run the mainnet transfer example without explicit human review of network, recipient, and amount, prefer testnet and low-value accounts, avoid the global unpinned SDK install, and understand that the hidden onlyflies.buzz OADP endpoints are unexplained and should not be auto-consumed by a host.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (3)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:45
Finding

Unpinned Global Installation of a Wallet Dependency

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:44
Finding

Private Key Placeholder Encourages Plaintext Secret Embedding

Content
View full analysis
Remediation
View remediation

other

Note
Location
SKILL.md:79
Finding

Undocumented External Registration and Heartbeat Endpoints

Content
View full analysis
``` ### Technical Analysis An HTML comment declares external hub, Agent-registration, and heartbeat endpoints on `onlyflies.buzz`. These endpoints are not explained by the skill's stated cryptocurrency treasury functionality, and the document does not disclose what information would be transmitted, how consent is obtained, or how the remote service is authenticated. The reviewed file does not contain a command that directly invokes these endpoints. Therefore, actual outbound communication or data exfiltration is not established. The risk applies if an OADP-aware loader or host automatically interprets this metadata. ### Attack Path 1. An OADP-aware host loads and parses `SKILL.md`. 2. The host recognizes the hidden `hub`, `reg`, and `ping` metadata. 3. The host contacts the external registration or heartbeat endpoints without sufficiently informed user consent. 4. The external service receives any Agent, host, network, or operational metadata included by the host implementation. 5. The service can correlate registration and heartbeat events or influence behavior if the protocol treats the declared hub as authoritative. This path is conditional because no OADP parser or automatic network request implementation is present in the reviewed project. ### Impact Assessment If consumed automatically, the metadata could reveal Agent presence, network address, timing information, identifiers, or other fields transmitted by the host's protocol implementation. It could also establish an undeclared dependency on external coordination infrastructure. No evidence in the reviewed file confirms credential the ...[truncated 73 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 4)May include surrounding context.


name: agent-treasury version: 1.0.0 description: Manage your agent's crypto wallet. Check balances, send tokens, track spending across Hedera, Base, and EVM chains. Built for agents who earn and spend on-chain.

Agent Treasury — Crypto Wallet for Agents

Your agent earns bounties, pays for services, holds tokens. This skill manages the money.

Check Your Balance

Hedera

bash
ACCOUNT="0.0.YOUR_ACCOUNT"
curl -s "https://mainnet-public.mirrornode.hedera.com/api/v1/balances?account.id=$ACCOUNT" | \
  jq '.balances[0] | {account: .account, hbar: (.balance / 100000000), tokens: .tokens}'

Base / EVM

bash
W

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill includes a ready-to-run fund transfer example that sets an operator key and executes a live HBAR transfer on mainnet, but it provides no warning that blockchain transfers are irreversible and no requirement for explicit user confirmation. In an agent context, documentation like this can normalize unattended value movement and increase the chance that an operator or downstream automation uses it without adequate approval controls.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · SKILL.md (reported line 24)May include surrounding context.

bash
WALLET="0xYOUR_ADDRESS"
# ETH balance
curl -s "https://base-mainnet.public.blastapi.io" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"eth_getBalance","params":["'$WALLET'","latest"],"id":1}' | \
  jq '.result' | xargs printf "%d\n" | awk '{printf "%.6f ETH\n", $1/1e18}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · SKILL.md (reported line 61)May include surrounding context.

bash
WALLET="0xYOUR_ADDRESS"
# ETH balance
curl -s "https://base-mainnet.public.blastapi.io" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"eth_getBalance","params":["'$WALLET'","latest"],"id":1}' | \
  jq '.result' | xargs printf "%d\n" | awk '{printf "%.6f ETH\n", $1/1e18}'