Back to skill

Security audit

Generate Fractional CFO Firm Client Education Handout

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to be a simple handout-writing helper, but it requests broad shell and local file access that its own workflow does not need.

Review this before installing. It does not contain malicious commands or persistence, but a handout generator should not normally need shell or broad file-read permissions; install only if you are comfortable with those permissions or the publisher removes them.

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

Unnecessary Shell and Local File Access Permissions

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:11
Vulnerability Type: Excessive tool permissions violating least privilege
Risk Level: Medium

Vulnerable Code:

yaml
allowed-tools: Bash, Read

Technical Analysis

The skill grants access to Bash and Read, but its documented workflow only requires generating and refining a fractional CFO client education handout. The declared APIs are chat and image_generation, and no documented workflow step requires shell command execution or arbitrary access to local files.

These permissions unnecessarily expand the skill's security boundary. Bash can provide broad command-execution capabilities under the agent process's operating-system identity, while Read may expose any local files accessible to that identity. Although the reviewed skill contains no direct malicious command or file-read instruction, excessive authorization increases the consequences of future instruction injection, compromised referenced content, or unsafe modifications to the skill.

Attack Path

  1. A user installs and loads the skill with Bash and Read enabled.
  2. The skill or content processed during a handout-generation session is later modified or contains attacker-controlled instructions.
  3. Those instructions induce the agent to invoke the unnecessarily authorized tools.
  4. The agent reads accessible local files or executes shell commands outside the legitimate handout-generation workflow.
  5. Retrieved information may be exposed through generated output, while shell commands may alter files or execute other processes within the agent account's permissions.

The current repository does not contain the malicious instructions required to complete this path; the finding concerns the avoidable privilege exposed if such instructions are introduced.

Impact Assessment

Successful exploitation could provide access to files readable by the agent's operating-system acco ...[truncated 468 chars]

Remediation
View remediation

Remediation Suggestions

  • Remove Bash and Read from allowed-tools because neither is required by the documented workflow.
  • Authorize only the capabilities necessary for content generation, such as the declared chat and image_generation APIs.
  • If file access is later required, replace arbitrary reads with narrowly scoped access limited to explicit user-selected input and output locations.
  • If command execution becomes necessary, expose specific fixed operations rather than a general-purpose shell, validate all arguments, and run them in a sandbox with minimal filesystem and network access.
  • Add a permission review to release checks so newly requested tools must be justified by a documented workflow step.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill provides a natural-language suggested prompt, "[Generate Fractional CFO Firm Client Education Handout] for my fractional CFO firm," but does not define clear activation boundaries, exclusions, or narrower trigger conditions. In a markdown skill file, this can create ambiguous invocation scope because the description relies on general-purpose phrasing like "for my fractional CFO firm" without negative examples or explicit constraints.

Content

No source excerpt is available for this finding.