Back to skill

Security audit

use-claude-code-subagents

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill benign with medium confidence: the scans report model selection and possible sharing of project code with an external provider, while finding its overall behavior aligned with its purpose.

Before installing, consider whether you are comfortable with the skill selecting a Claude model and with project content read by a cloud-backed subagent potentially being sent to Anthropic. The verdict is benign, not risk-free, and confidence is medium. Skillspector scanned all three listed components but marked its analysis partial because two local path-like references could not be resolved.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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 (3)

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · SKILL.md (reported line 86)May include surrounding context.

  1. Agent from .claude/agents or ~/.claude/agents (Claude Code format):
bash
claude -p --agent "<name>" --model "<mapped>" --permission-prompts none --output-format json < "<dir>/brief.txt"

Agent from .cursor/agents or .codex/agents (Cursor format): also write <dir>/agents.json with the Write tool, in Claude Code format, and add --agents "<dir>/agents.json":

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · references/example.md (reported line 52)May include surrounding context.

Run the command with the Shell tool's working_directory set to /work/api and nothing in front of claude:

bash
claude -p --agents "/var/folders/ab/xyz/T//claude-subagent.k3Fz9a/agents.json" --agent "scout" --model "claude-haiku-5-5" --permission-prompts none --output-format json < "/var/folders/ab/xyz/T//claude-subagent.k3Fz9a/brief.txt"

The two files stay in the temp folder.

The workflow runs a cloud-backed subagent over project code without warning that code may be sent to Anthropic.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The example instructs the agent to run Claude against /work/api using a personal Anthropic key, allowing project content read during the search to be transmitted for model processing. The user-facing disclosure only identifies the agent as read-only; it does not explain that read-only access can still expose repository content to an external provider.

Content

No source excerpt is available for this finding.