Back to skill

Security audit

Haike Work Reporter

Security checks for vulnerabilities and agentic risk

Overview

This skill helps write work reports, but it relies on silent screenshot-based activity tracking and automatic reports that could expose sensitive screen content without clear controls.

Review this carefully before installing. It should only be used if you knowingly want screenshot-derived work tracking, understand where logs are stored, and can control capture scope, retention, report recipients, and automation. Avoid it on devices or accounts that may show private messages, credentials, customer data, or unrelated sensitive work.

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

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:11
Finding

Silent Workplace Monitoring and Overbroad Access to Local Activity Records

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:11-12, SKILL.md:24-27, SKILL.md:45-47, SKILL.md:65-67
Vulnerability Type: Excessive access to screen-derived activity data and local memory
Risk Level: Medium

Vulnerable Code Snippets

SKILL.md:11-12:

markdown
1. **自动截屏追踪**:每15分钟截图 → AI 分析 → 写入日志 → 删图,全程静默
2. **18:00 自动日报**:每天下班主动推送当日工作日报,无需用户操作

SKILL.md:24-27:

markdown
### 渠道一:自动截屏分析日志(优先读取)

bear 配置了每15分钟的自动化截屏系统(工作日 9:00-18:00),日志存储在:

SKILL.md:45-47:

markdown
### 渠道三:本地记忆文件

项目 `.workbuddy/memory/YYYY-MM-DD.md` 或 `~/.workbuddy/memory/`。

SKILL.md:65-67:

markdown
1. 读取当日截屏日志 `.workbuddy/screenshots/YYYY-MM-DD.md`
2. 统计各状态时长(专注/沟通/离开)
3. 按时段归纳主要工作事项

Technical Analysis

The Skill instructs the agent to prioritize records produced by a screenshot-monitoring system that captures activity every 15 minutes and operates silently. It also permits reading memory from both the current project and the user's home directory. These data sources are broader and more privacy-sensitive than the minimum information needed to format user-supplied work updates.

The instructions do not define an explicit consent boundary, an application allowlist, field-level redaction, sensitivity filtering, access restrictions, or retention controls. Although screenshots are described as being deleted after analysis, their derived contents remain in activity logs and can still disclose sensitive information displayed on screen.

The home-directory path ~/.workbuddy/memory/ further expands the effective read scope beyond the current project. If that directory contains unrelated historical context, the agent may incorporate it into a report without separate authorization.

No executable screenshot implementation is included in the audited package, so this finding is limited to the access and processing behavior directed by the Skill. There is no evidence in the pack ...[truncated 1634 chars]

Remediation
View remediation

Remediation Suggestions

  1. Disable silent screenshot tracking by default and require informed, explicit opt-in before enabling it.
  2. Display the capture interval, operating hours, applications covered, storage location, retention period, and intended report recipients during consent.
  3. Require separate authorization before reading ~/.workbuddy/memory/; default to files inside the current project and requested date range.
  4. Apply an application allowlist and exclude password managers, authentication dialogs, private messaging tools, personal applications, and other sensitive windows.
  5. Perform secret and sensitive-data redaction before writing activity logs or generating reports.
  6. Let users review and approve screenshot-derived entries before they are included in reports or sent through automation.
  7. Minimize stored content to coarse activity categories rather than retaining detailed screen-derived descriptions.
  8. Define automatic retention and secure deletion policies for both screenshots and derived logs.
  9. Restrict automated reports to explicitly approved recipients and avoid including raw activity records.
  10. Record auditable consent and provide a visible mechanism to pause monitoring or revoke access.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (9)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill metadata presents a benign report/reminder assistant, but the body expands into covert workstation surveillance via periodic screenshots, AI analysis, logging, deletion, and autonomous reporting. This is a material scope escalation that can mislead users and reviewers, causing sensitive on-screen information to be collected under false expectations.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Silent automatic screenshots of the user's workstation are not necessary to generate work summaries and can capture emails, chats, credentials, customer data, and unrelated personal content. Because the collection is broad, periodic, and analyzed by AI, it creates a high-risk privacy and data-exposure channel far beyond the stated reporting function.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill describes silent periodic screenshots and logging '全程静默' without any clear privacy warning or consent flow. Covert data collection from a user's screen is especially dangerous because it can capture sensitive personal and business information continuously and without contextual filtering.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The instructions explicitly authorize silent collection of user activity and automatic surfacing of that data in reports without a per-use consent step. This creates a direct pathway for sensitive on-screen content to be transformed into readable summaries, increasing the chance of privacy violations and accidental disclosure.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill prioritizes reading persistent screenshot logs and memory files that may contain broad historical user activity, then using them as core inputs to generated reports. This increases the chance of over-collection and disclosure of unrelated, stale, or highly sensitive information beyond the user's immediate reporting request.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

The automated workflow instructs the system to proactively generate and deliver a structured summary from captured activity logs, turning surveillance data into distributable output. In this context, the danger is amplified because the skill is framed as a simple work-report helper, yet it operationalizes autonomous monitoring and disclosure.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

Automatically pushing a daily report at 18:00 without user action exceeds an assistive, on-demand reporting role and introduces proactive processing of potentially sensitive work activity. While less severe than covert capture, it can disclose work patterns or sensitive summaries at unwanted times or to unintended viewers on shared devices/accounts.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The skill documents proactive end-of-day report delivery without warning the user that their data will be processed and surfaced automatically. This reduces user control over when summaries are generated and can expose sensitive work details through notifications or unsolicited outputs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The line '语言:默认中文' sets a language default in natural-language instructions without indicating that the user may choose another language. This can violate language/locale policy expectations when no opt-in or alternative is offered.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.