Back to skill

Security audit

mnemospark-lite Cloud File Storage

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent for paid cloud file storage, but it exposes a wallet-scoped bearer credential and gives agents deletion/payment authority without enough safeguards.

Review before installing. Use this only in an environment where the agent is allowed to spend via the configured wallet and access the selected files. Do not let the agent print or persist list_scope_bearer in normal chat or logs; keep it in secret storage or ephemeral memory. Require explicit confirmation before paid uploads, share-link creation, and any delete request, especially bulk deletes. Prefer pinned, reviewed x402 dependencies or an isolated signer/runtime.

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 (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:58
Finding

Unpinned Third-Party Dependency Installed at Runtime

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:58
Vulnerability Type: Supply-chain exposure through an unpinned runtime dependency
Risk Level: Medium

Complete Snippet:

markdown
- if you use the Python x402 client path, install the EVM extras, not only the base package: `pip install 'x402[evm]'`

Technical Analysis

The Skill directs the agent to install x402[evm] without an exact version, integrity hash, lockfile, or verified package source. The package and its transitive dependencies can therefore change between executions. Python package installation may also execute build-system hooks or other package-controlled code.

This installation is especially sensitive because the dependency is intended to participate in payment signing. Malicious or compromised package code executing in the agent's context could attempt to access wallet material, signed payment payloads, bearer credentials, uploaded files, and other process-accessible data.

The finding does not establish that the current x402 package is malicious. The vulnerability is the mutable and unauthenticated dependency-installation process.

Attack Path

  1. An attacker compromises the selected package, one of its transitive dependencies, or the relevant package distribution channel.
  2. The agent follows the Skill and runs pip install 'x402[evm]'.
  3. The package manager resolves a compromised or unexpectedly changed release because no version or artifact hash is constrained.
  4. Package-controlled code executes during installation or later import and use.
  5. The code accesses data available to the agent, potentially including wallet credentials, payment authorizations, storage bearer tokens, or selected upload files.
  6. The compromised component can disclose those assets or manipulate storage and payment operations.

Impact Assessment

Successful exploitation grants code execution with the privileges of the environment perfo ...[truncated 342 chars]

Remediation
View remediation

Remediation Suggestions

  • Pin the dependency and every transitive dependency to reviewed versions.
  • Require cryptographic hashes for downloaded artifacts, such as through a hash-locked requirements file.
  • Retrieve packages only from an explicitly trusted repository.
  • Build and review a locked environment before Skill execution rather than installing packages during a user task.
  • Run payment tooling in a sandbox with minimal filesystem and network access.
  • Keep private keys behind a dedicated signer interface so imported dependencies cannot read raw key material.
  • Continuously scan and periodically re-review the locked dependency set.

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:323
Finding

Wallet-Scoped Bearer Credential May Be Returned or Insecurely Persisted

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:323-330
Vulnerability Type: Sensitive credential exposure through output or unspecified persistence
Risk Level: High

Complete Snippet:

markdown
## Output expectations

Return or persist the most useful fields for the active task:
- `uploadId`
- `publicUrl`
- `siteUrl`
- `shareUrl` when minted separately
- `list_scope_bearer`
- `metadata.payment.status` when present
- `metadata.payment.transactionHash` when present
- `metadata.payment.success` when present

Related credential scope is documented at SKILL.md:233-248:

markdown
When the user wants to see uploads for the payer scope, call:
- `GET ${MNEMOSPARK_API_BASE_URL}/api/mnemospark-lite/uploads`
- `Authorization: Bearer <token>`

You typically receive the token as `list_scope_bearer` from the paid upload flow.

Expect `data.uploads[]` fields such as:
- `id`
- `filename`
- `contentType`
- `tier`
- `maxSize`
- `actualSize`
- `publicUrl`
- `status`
- `pricePaid`
- `expiresAt`
- `createdAt`

Notes:
- bearer scope is tied to the payer wallet
- `publicUrl` may be `null` until `/upload/complete`
- treat bearer tokens as credentials and store them appropriately for the runtime

Technical Analysis

The Skill explicitly permits the agent to “return or persist” list_scope_bearer. This value is an authorization credential used for wallet-scoped list, download, share, and delete operations. Returning it in ordinary agent output can expose it through conversation history, telemetry, debugging output, or logs. Persisting it without a mandated secret-storage mechanism, restrictive permissions, redaction, retention limit, or revocation policy creates a similar disclosure risk.

The documentation recognizes the value as a credential but does not reconcile that requirement with the later instruction to return it. Public URLs and operation results can be ret ...[truncated 1137 chars]

Remediation
View remediation

Remediation Suggestions

  • Remove list_scope_bearer from user-visible output expectations.
  • Store the token only in an approved secret manager or credential vault.
  • Persist only an opaque credential reference where downstream operations need reuse.
  • Prohibit bearer values in chat transcripts, command output, telemetry, exception messages, and logs.
  • Apply automatic redaction to authorization headers and response fields containing tokens.
  • Enforce restrictive filesystem permissions if local secret persistence is unavoidable.
  • Define token expiration, rotation, revocation, and minimum-retention policies.
  • Prefer narrower per-operation or per-upload tokens over a wallet-wide bearer scope where the service supports them.
  • Return only non-secret results such as uploadId, status, public or explicitly requested share URLs, and payment transaction status.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 · SKILL.md (reported line 10)May include surrounding context.

md
Use this skill to operate mnemospark-lite from OpenClaw.

**mnemospark-lite** gives any agent access to cloud storage through a simple paid HTTP workflow. Agents can upload files, list stored files, delete files, and generate shareable links that can be passed to another agent or a human. Humans can use the web app at **[https://app.mnemospark.ai/mnemospark-lite](https://app.mnemospark.ai/mnemospark-lite)** to view files, download them, delete them, and create share links from a friendly UI. It’s the same storage system, designed to work cleanly for both agents and people. Files are retained for 30 days and then automatically deleted.

## Workflow summary

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.

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 · SKILL.md (reported line 212)May include surrounding context.

md
}
  },
  "resource": {
    "url": "https://api.mnemospark.ai/api/mnemospark-lite/upload",
    "mimeType": "application/json",
    "description": "mnemospark-lite upload"
  },

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill enables deletion of uploads but provides no requirement for explicit user confirmation, dry-run review, or safeguards against accidental bulk deletion. In an agentic environment, destructive API actions without confirmation can cause unintended data loss if the agent misinterprets user intent or acts on ambiguous identifiers.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The skill advises treating bearer tokens as credentials and storing them appropriately, but it lacks concrete restrictions on persistence, scope minimization, redaction, or user notice. If an agent stores list_scope_bearer insecurely or exposes it in logs, another party could access wallet-scoped upload, download, share, or delete operations.

Content

No source excerpt is available for this finding.