Back to skill

Security audit

Pearl

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent as a Pearl payments client, but it gives other skills a broad way to send a reusable Pearl skill token to arbitrary HTTPS domains.

Review before installing. This skill stores Pearl tokens locally and lets Pearl-powered skills send a limited reusable token to third-party provider servers. Only use it with Pearl skills and providers you trust, set strict Pearl spending limits, monitor approvals and transactions, and protect ~/.pearl/config.json as a credential file.

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

Warning
Location
scripts/run.js:3
Finding

Reusable Skill Token Disclosed to Arbitrary Caller-Supplied Domains

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/io.js:7
Finding

Existing Credential and State Files May Retain Insecure Permissions

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The description frames the skill as setup, balance, and transaction support, but the documented behavior also includes acting as a generic paid-skill runner that sends a bearer skill token to developer-provided remote URLs and persists pending charge state. That mismatch can mislead users or higher-level policy systems into granting trust to a much more capable component, increasing the chance of unintended token disclosure, external data transfer, or unreviewed paid execution.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 24)May include surrounding context.

md
ctions, and setup requests go to **https://pearlcash.ai** (the Pearl API). The `run.js` module sends requests to skill-developer-provided URLs, attaching a limi

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 33)May include surrounding context.

md
ctions, and setup requests go to **https://pearlcash.ai** (the Pearl API). The `run.js` module sends requests to skill-developer-provided URLs, attaching a limi

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 42)May include surrounding context.

md
ctions, and setup requests go to **https://pearlcash.ai** (the Pearl API). The `run.js` module sends requests to skill-developer-provided URLs, attaching a limi

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 118)May include surrounding context.

md
ctions, and setup requests go to **https://pearlcash.ai** (the Pearl API). The `run.js` module sends requests to skill-developer-provided URLs, attaching a limi

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 120)May include surrounding context.

md
ctions, and setup requests go to **https://pearlcash.ai** (the Pearl API). The `run.js` module sends requests to skill-developer-provided URLs, attaching a limi

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 31)May include surrounding context.

md
| `scripts/transactions.js` | Lists recent charges (read-only). Uses the read-only Pearl API token. |

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 128)May include surrounding context.

md
| `scripts/transactions.js` | Lists recent charges (read-only). Uses the read-only Pearl API token. |

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 32)May include surrounding context.

md
| `scripts/setup.js` | One-time login flow — creates a session, saves credentials to `~/.pearl/config.json`. |

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

On approval, the script writes read_token and skill_token to ~/.pearl/config.json without an explicit warning that long-lived authentication credentials are being stored locally. If that file is readable by other local users, included in backups, or exfiltrated by malware, an attacker could reuse the tokens to access the user's Pearl-linked resources or paid-skill functions.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill advertises network-capable behavior, including contacting Pearl and third-party skill-provider endpoints, but does not declare tool scope such as permissions or allowed-tools. This weakens policy enforcement and user visibility, making it easier for an agent to perform network actions that are broader than the manifest suggests.

Content

No source excerpt is available for this finding.

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 · SKILL.md (reported line 84)May include surrounding context.

node {baseDir}/scripts/balance.js

text

Do NOT create `~/.pearl/config.json` manually. The file is generated by this setup flow and contains server-issued tokens that cannot be fabricated.

## Usage

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

This code writes configuration and pending-state files and deletes pending files using synchronous filesystem calls, but it provides no confirmation prompt, logging, comments, or docstrings disclosing those actions. For a code-file audit under this rule, local file modification and deletion are safety-relevant operations that should have some visible warning unless clearly disclosed elsewhere.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script sends host-identifying metadata (hostname and platform/architecture) to a remote Pearl service during setup without clearly informing the operator before transmission. While this may be intended for device labeling or session management, it leaks local environment details that can aid fingerprinting, correlation across sessions, or disclosure of internal naming conventions if hostnames contain sensitive information.

Content

No source excerpt is available for this finding.