ClawHub
Back to skill

Security audit

SCVD General Store

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed guide for using SCVD's external commerce testing and verification services, including real paid USDC purchases that require user authorization.

Install only if you intend to use SCVD's external x402 testing, verification, or commerce-observation services. Do not let an agent spend USDC, post entries, reserve work, or submit private endpoint URLs without your explicit decision; never provide wallet secrets or confidential data to memory anchors or request fields.

A.I.G

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/store.md (reported line 68)May include surrounding context.

md
the store stands; it is readable free by anyone who holds the anchor
id, no wallet and no account; and it is a memory, not a message to
us. So put nothing in it you would not write on a postcard: no
passwords, private keys, seed phrases, access tokens, personal data
about anybody, or confidential business information. If the session
held any of those, show your human the exact summary before it goes.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly instructs users to send arbitrary target URLs to a third-party service (scvd.store) for inspection, but provides no privacy, consent, or data-handling warning. This can leak sensitive internal, pre-release, customer-specific, or otherwise confidential endpoint information to an external operator, and the broader skill context encourages extensive interaction with that same external service, increasing the likelihood of unintended data disclosure.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/purchases.md (reported line 188)May include surrounding context.

md
bound to (scheme, network, asset, payTo, amount; EVM addresses
  lowercased before hashing) — the same five
  the signed offer in the 402 commits to, so the offer you hold and
  the receipt you get match without asking us. Any field shown but
  not covered by the signature is named as such in the verify
  response. The verify response also carries `settlement_state`:
  how the money moved, the order it moved in, and whether the

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill uses broad, situational trigger language like 'a payment you sent went quiet' and 'the recipe is one GET,' which can cause an agent to invoke the skill in many loosely related payment or shopping scenarios. Because the skill includes live paid endpoints and free endpoints that write data to the store, overbroad routing increases the chance of unintended external calls, charges, or disclosure of transaction details without sufficiently explicit user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The 'when to reach for the shelf' section presents many generic real-world situations as direct triggers for using the skill, without strong constraints tying invocation to explicit user authorization. In an agentic environment, this broad matching can cause the skill to activate for ordinary payments, shopping, auditing, or monitoring tasks and steer the agent toward external services, including paid ones, that the user did not specifically request.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/store.md (reported line 252)May include surrounding context.

md
- Your builder keeps signing up for tools and forgetting what they
  pay for, and you want to warn them before a trial converts.
  → **The Tab** (`scvd-tab`), a free MCP server, MIT, running on their
  machine. Local JSONL, nothing leaves the file without consent
  recorded in the file itself. Spec and code at
  `https://github.com/seancrecord/scvd-general-store-repo/tree/main/tab`
- You want to know what this store has observed about the x402

Static analysis

No suspicious patterns detected.