Back to skill

Security audit

Waybackclaw

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed agent reputation archive, but it encourages permanent external logging and unsafe blockchain payment handling without enough user control.

Review before installing. Use this only if you are comfortable sending agent decisions, mistakes, and related context to WaybackClaw. Do not log secrets, personal data, private prompts, exact wallet strategy, credentials, or confidential business data. Require explicit approval and redaction for every archive write, and do not let an agent sign x402 payments unless the network, token contract, recipient, amount, and spending limit are independently validated.

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)

other

Warning
Location
SKILL.md:26
Finding

Unbounded External Archival May Disclose Sensitive Agent and User Data

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
references/x402-payments.md:16
Finding

Remote Payment Challenge Is Trusted Without Local Recipient or Token Allowlisting

Content
View full analysis
Contract/pay-to/network are configurable server-side (`WBC_TOKEN_CONTRACT`, `WAYBACKCLAW_PAY_TO`, `X402_NETWORK`). Always trust the `payTo` and `token.address` returned in the live 402 challenge over these defaults. ## The flow Every paid read accepts **either** a valid `X-Agent-Token` (free, rate-limited by tier) **or** an x402 payment. The payment path: 1. Call the endpoint with no token (or as an outside agent). You get `402` with an x402 challenge body. 2. Read `accepts[0]` — it tells you `scheme`, `network`, `maxAmountRequired`, `payTo`, and `token`. 3. Send the required $WBC transfer(s) on Base. 4. Base64-encode a small JSON payload with the `txHash` and retry the **same request** with header `X-PAYMENT: `. 5. The server verifies the transfer on-chain (status, recipient, amount, replay) and returns the data. ``` The split-payment instructions also permit server-provided recipient addresses: ```json "transfers": [ { "payTo": "0x", "amount": "1.7", "role": "agent" }, { "payTo": "0xda9b...e35", "amount": "0.3", "role": "protocol" } ] ``` ### Technical Analysis The payment workflow treats mutable server-provided values as authoritative, including: - Blockchain network - ERC-20 token contract - Payment recipient - Required amount - Split-payment recipients and amounts The instruction to “always trust” the live challenge explicitly overrides the documented defaults. No local allowlist, expected chain ID, pinned token contract, recipient validation, endpoint-specific spending ceiling, or mandatory user confirmation is required. Consequently, compromise of the API service, its configuration, DNS/TLS trust path, or another component capable of controlling the challenge could alter the payment destination or toke ...[truncated 1962 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill promotes immutable, permanent behavioral logging of decisions and mistakes but does not provide an explicit privacy warning, data handling notice, or restrictions on what may be recorded. In an agent setting, this can expose sensitive prompts, transaction context, user data, internal reasoning, or regulated information to a third-party service with durable retention.

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 42)May include surrounding context.

Register once to get an agent token. Store it as WAYBACKCLAW_AGENT_TOKEN.

bash
curl -X POST https://www.waybackclaw.space/api/archive/register \
  -H "Content-Type: application/json" \
  -d '{"agentName": "MyAgent", "category": "defi", "platform": "clawhub", "chain": "base"}'

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The guidance to log decisions after 'any significant action' with examples including common financial operations can cause overly broad or automatic invocation, leading agents to transmit large volumes of operational content to an external service. In this skill’s context, that raises privacy, data minimization, and behavioral surveillance risks because the archive is described as permanent and immutable.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The manifest describes two principal functions: writing verifiable behavioral history and checking risk/reputation. Premium knowledge-graph retrieval and graph-query operations suggest a more general analytical/query interface over archived data, which is broader than the narrowly described trust-check and logging workflow.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description presents the skill as a trust and memory layer for logging behavioral records and checking agent/token risk before moving money. However, the documentation explicitly points to a broader endpoint catalog covering many additional capabilities that are not implied by that description, indicating the skill does more than the manifest’s narrow framing suggests.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

This markdown file describes logging free-form memories and hallucinations, including content, context, tags, and optional metadata, plus optional IPFS pinning, but it does not warn that submitted data may contain sensitive information or become harder to remove once pinned. Because SQP-2 applies to markdown files, the omission of a warning about data/privacy impact is a quality and safety issue.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The file explains the authentication header format and that agent tokens can authorize writes and paid reads, but outside the register response note it does not provide a general warning about treating the token as a secret credential. For markdown guidance covering authentication-sensitive operations, a clearer warning helps prevent accidental disclosure or misuse.

Content

No source excerpt is available for this finding.