Back to skill

Security audit

Clipboard History

Security checks for vulnerabilities and agentic risk

Overview

This clipboard-history skill is not clearly malicious, but it would persist sensitive clipboard contents and has noisy, overbroad instructions that should be reviewed before use.

Review before installing. Only use this skill if you are comfortable with copied text being retained locally, including accidental copies of passwords, tokens, private documents, or personal data. Prefer a version that requires explicit opt-in, encrypts or restricts the history file, supports exclusions and one-click purge, and removes the unrelated merged trigger/content sections.

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 (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:13
Finding

Unprotected Persistent Storage of Clipboard Contents

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 13–17
Vulnerability Type: Plaintext storage of potentially sensitive clipboard data
Risk Level: Medium

Vulnerable Code Snippet

markdown
## Usage
- Automatically saves every copied item (latest 50 entries)
- Supports viewing, searching, and restoring history
- Data is stored locally and is not transmitted over the network

## Data File
`D:\coze-local\data\clipboard_history.json`

The snippet above is an English rendering of the behavior documented in the original file.

Technical Analysis

The Skill declares that it automatically records the latest 50 clipboard entries in a predictable local JSON file. Clipboard contents may include passwords, authentication tokens, private messages, personal information, recovery codes, and confidential business data.

The specification does not describe encryption at rest, restrictive file permissions, sensitive-content filtering, explicit capture consent, configurable expiration, or secure deletion. Consequently, sensitive values may remain recoverable after they are no longer present on the active clipboard. Although the document says that data is not transmitted over the network, local-only storage does not prevent disclosure to another local user or process with read access to the file.

The repository contains only SKILL.md; therefore, the audit could not verify the actual storage format or runtime access controls. This finding is based on the explicitly documented automatic persistence behavior.

Attack Path

  1. The user enables or invokes the clipboard-history Skill.
  2. The user copies sensitive content, such as a password, token, recovery code, or confidential text.
  3. The Skill automatically adds that content to its retained history.
  4. The entry is stored at the predictable path D:\coze-local\data\clipboard_history.json.
  5. A local user, malware process, backup utility, or other application with access to that file reads ...[truncated 765 chars]
Remediation
View remediation

Remediation Suggestions

  1. Require explicit, informed opt-in before monitoring or retaining clipboard contents.
  2. Clearly warn users that copied secrets may be persisted beyond the lifetime of the active clipboard.
  3. Encrypt clipboard history at rest using a key protected by the operating system's credential store.
  4. Create the data file with permissions restricted to the current user and reject insecure permission configurations.
  5. Filter or avoid retaining likely secrets, including passwords, authentication tokens, private keys, recovery codes, and payment information.
  6. Integrate with platform APIs that identify password or sensitive clipboard content where available.
  7. Provide configurable retention limits, short expiration periods, per-entry deletion, and a one-action history purge.
  8. Implement secure deletion where supported and document the limitations introduced by backups, journaling, and solid-state storage.
  9. Avoid a globally predictable or shared storage location; use an application-specific directory scoped to the current user.
  10. Add automated tests verifying encryption, access permissions, expiration, exclusion rules, and deletion behavior.
  11. Document the exact data format, threat model, access controls, and lifecycle before distributing the Skill.
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 (8)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Automatically storing clipboard contents without a prominent warning is dangerous because clipboards often contain passwords, tokens, personal data, internal documents, or financial information. In this skill’s context, the feature is explicitly designed to capture every copy event, so the privacy and secret-exposure risk is inherent and materially elevated.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill claims clipboard history is stored locally and 'does not use the network', yet the documentation includes external Bilibili references and unrelated learned/merged content. Even if these URLs are only documentation, this inconsistency undermines trust, suggests uncontrolled content ingestion, and can mislead reviewers or users about the skill’s actual data flows and provenance.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad everyday terms such as 'clipboard' and 'copy history', making accidental activation likely during ordinary conversation. In a clipboard-history skill, unintended activation is more dangerous than usual because it can expose or manipulate highly sensitive copied content without the user clearly intending to invoke the feature.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Large portions of the skill file are unrelated to clipboard history and contain merged Bilibili notes about music, percussion, and other irrelevant topics. This bloated, mixed-purpose content increases prompt-injection and unintended-trigger risk because the agent may ingest and act on irrelevant instructions or noisy phrases that were never meant to define clipboard functionality.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This later trigger block introduces ambiguous and unrelated activation phrases, including terms disconnected from clipboard management. Such polluted trigger definitions can cause the wrong skill to activate, or activate in contexts shaped by unrelated merged content, increasing the risk of unintended disclosure of clipboard history.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The duplicated trigger section repeats vague and irrelevant activation conditions, compounding the chance of false activation. Repetition across merged sections suggests weak configuration hygiene, which is risky for a skill that can surface sensitive clipboard contents on casual mention of overlapping terms.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

Another repeated trigger section continues the pattern of vague or unrelated activators, indicating that merged content has contaminated the skill’s activation logic. In this context, each extra ambiguous trigger increases the chance of accidental execution and unauthorized exposure of stored clipboard entries.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The visible user-facing description and usage instructions are presented only in Chinese, with no indication that the skill supports other languages or that Chinese is a required locale. For a general-purpose clipboard utility, restricting the interaction language without opt-in can violate language-choice policy expectations.

Content

No source excerpt is available for this finding.