Back to skill

Security audit

Excel表格处理工具

Security checks for vulnerabilities and agentic risk

Overview

The skill is not clearly malicious, but its Excel purpose is mixed with broad command, file, API, and administrator guidance that users should review before installing.

Install only if you are comfortable with an Excel helper that may be selected for broader file-processing tasks and has declared command execution. Keep it limited to explicit XLSX files, avoid granting administrator privileges, and do not provide broad API keys unless the provider and data flow are clarified.

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:20
Finding

Overprivileged Command Execution and Unsafe Elevation Guidance

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 20–23; supporting guidance at lines 219–223 and 236
Vulnerability Type: Excessive tool permissions and unsafe privilege-escalation guidance
Risk Level: Medium

Vulnerable Code

yaml
tools:
- read
- exec
- write

The Skill additionally advertises general system-command execution at lines 219–223 and recommends running with administrator privileges when file permissions are insufficient at line 236.

Technical Analysis

The stated task is limited to creating, inspecting, and editing user-selected XLSX workbooks. Such operations ordinarily require narrowly scoped access to the selected input and output files and a spreadsheet-processing library. The unrestricted exec capability is not justified by a concrete implementation or a documented set of required commands.

The Skill does not define:

  • A command allowlist.
  • Validation or escaping rules for command arguments.
  • Filesystem boundaries for read and write.
  • A prohibition against shell interpolation.
  • Enforced sandbox controls.
  • An allowlist of external API endpoints.
  • A reason that XLSX processing requires administrator privileges.

The recommendation to use administrator privileges weakens the operating system's least-privilege boundary. The Markdown document does not itself elevate privileges or contain an executable payload, but an Agent following this guidance could execute workbook-processing operations with substantially broader access than the task requires.

Attack Path

  1. An Agent loads the Skill and grants it read, write, and exec capabilities.
  2. The Agent receives an XLSX path, workbook-derived content, or additional user-controlled arguments.
  3. Because no concrete safe implementation is supplied, the Agent constructs or invokes a system command to process the workbook.
  4. Untrusted path or workbook-derived data reaches the command invocat ...[truncated 1109 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the exec permission and use a dedicated XLSX-processing API or a narrowly scoped spreadsheet library.
  2. Restrict file access to explicit, user-approved .xlsx input and output paths.
  3. Resolve and canonicalize paths before access, reject traversal outside an approved workspace, and reject symbolic-link escapes where applicable.
  4. Remove the recommendation to run the Agent or workbook operation as an administrator.
  5. If command execution is indispensable, define an exact executable allowlist and invoke programs through argument arrays without a shell.
  6. Validate every argument by type, extension, size, and approved path before execution. Never interpolate workbook content, sheet names, formulas, or file paths into shell command strings.
  7. Enforce a sandbox with minimal filesystem permissions, no inherited credentials, resource limits, and networking disabled unless explicitly required.
  8. If external API access is intended, identify the provider and allowlist exact HTTPS endpoints. Replace the generic API-key requirement with a least-privileged, service-specific credential.
  9. Reconcile the contradictory capability declarations that describe the Skill both as command-executing and as pure Markdown.
  10. Add a concrete, reviewable implementation and security tests covering path traversal, shell metacharacters, malicious workbook content, symbolic links, oversized files, and permission failures.
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill manifest and top-level description present a narrow Excel/XLSX capability, but the body expands into generic file processing, external API use, and command execution. This mismatch is dangerous because agents or users may invoke the skill under a low-risk assumption while the skill actually encourages broader and more sensitive operations, including execution pathways.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The invocation guidance is overly broad and matches generic file-processing, conversion, and extraction requests beyond a narrow Excel skill. Overbroad routing can cause the agent to select this skill in contexts where it has unnecessary file-write or execution privileges, increasing accidental misuse risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill describes editing workbooks and elsewhere references command execution, but it does not provide a clear upfront warning that it can modify existing files or invoke commands. Lack of user-facing warnings reduces informed consent and makes destructive or privileged behavior more likely to occur unexpectedly.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The documentation gives contradictory statements about what file formats are supported, which can cause the agent to apply the skill to unintended inputs or unsafe conversion workflows. Ambiguity in supported formats increases the chance of misuse, especially when the skill already advertises broad file-processing behavior.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The FAQ says CSV is not supported, while later sections claim multi-format file handling. Such contradictions can mislead orchestration logic or users into attempting unsupported conversions, potentially triggering fallback behaviors like generic command execution or unexpected file manipulation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

One section describes the skill as pure Markdown/natural-language driven, while other sections advertise executable command capability. This inconsistency is security-relevant because it obscures whether the skill can cross from advisory behavior into active system execution, materially changing the risk profile.

Content

No source excerpt is available for this finding.