Back to skill

Security audit

Hotnews Excel

Security checks for vulnerabilities and agentic risk

Overview

The skill has a coherent Excel-processing purpose, but it also instructs persistent daily cron execution of an unaudited script outside the package.

Install only if you intentionally want recurring automated processing and you can verify and control /workspace/scripts/read_hotnews.py. Before use, require explicit opt-in for cron, define where outputs and logs are retained, and provide a clear way to disable the scheduled task and delete generated files.

Vulnerability Patterns
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • 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)

T06 · System Persistence

Error
Location
SKILL.md:69
Finding

Persistent Scheduled Execution of an Unverified External Script

Content
View full analysis
> /workspace/logs/hotnews.log 2>&1 ``` ``` ### Technical Analysis The Skill directs the environment to configure a recurring cron job that executes `/workspace/scripts/read_hotnews.py` every day. Scheduled tasks survive the original Skill invocation and create a cross-session execution mechanism, matching `T06: System Persistence`. The referenced Python script is not included in the audited project, which contains only `SKILL.md`. Its implementation, integrity, ownership, and write permissions therefore cannot be verified. Because the cron entry references a mutable external path, any party capable of creating or replacing `/workspace/scripts/read_hotnews.py` could cause arbitrary Python code to execute under the cron owner's account. The redirection also appends output to `/workspace/logs/hotnews.log`, allowing repeated execution to continue without interactive visibility. The cron expression itself does not establish which user installs it, so the exact privileges depend on the account under which the scheduled task is configured. ### Attack Path 1. A user activates the Skill while processing a matching Excel file. 2. The Skill's scheduling instruction is followed and the cron entry is installed. 3. The scheduled task remains active after the original processing session ends. 4. An attacker who can create or modify `/workspace/scripts/read_hotnews.py` places attacker-controlled Python code at that path. 5. At 08:00, cron changes the working directory to `/workspace` and executes the modified script. 6. The attacker's code runs with the ...[truncated 666 chars]
Remediation
View remediation
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 (5)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger words 'skill', '优化', '数据', and 'data' are extremely broad and likely to appear in ordinary conversation, which can cause unintended activation. In a skill that reads files, writes JSON, and may run follow-on processing, accidental invocation increases the chance of unreviewed data handling and user surprise.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest description is materially incomplete compared to the body of the skill: it omits local file persistence, JSON output paths, and automation behavior. This can mislead users or platform policy enforcement into approving a skill with broader data handling than disclosed, reducing informed consent and oversight.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill performs persistent writes to /workspace/data and describes scheduled processing, but it does not present a clear user warning or consent step about storage, logs, or automation. That is dangerous because users may believe the action is ephemeral when the skill actually creates retained artifacts and potentially recurring processing behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The cron configuration introduces recurring autonomous execution even though the skill is described as acting when a user sends a file. Undisclosed scheduled processing expands the trust boundary, may process data without a contemporaneous user action, and can create ongoing data collection or retention risks.

Content

No source excerpt is available for this finding.