Back to skill

Security audit

invest_skill_copilot

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because it reports a mismatch between the advertised research and report features and the supplied execution path, alongside a concern about broad, undeclared tool capabilities.

Consider the verdict’s DO_NOT_INSTALL recommendation. Before installing, seek a reviewed version that implements its advertised features, declares an appropriate tool scope, and addresses the reported synchronization and subprocess concerns.

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
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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
Findings (4)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code is a registration stub rather than an implementation of the declared research workflow. Comments attribute real execution to SKILL.md and scripts/repo_sync.py, but neither file is supplied, and run() does not invoke or delegate to them. Consequently, this code's observable behavior materially falls short of the declared primary purpose. No undeclared resource access or unrelated capability is evident.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/repo_sync.py (reported line 88)May include surrounding context.

python
def run(cmd, cwd=None, timeout=60):
    p = subprocess.run(cmd, cwd=cwd, capture_output=True, text=True, timeout=timeout)
    return p.returncode, p.stdout.strip(), p.stderr.strip()

Manifest parsing failures are silently ignored and snapshots are overwritten, contradicting the documented preserve-on-failure policy.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

Manifest parsing and schema-access exceptions are silently swallowed, after which the affected skill is still appended with default or partially assigned metadata. Synchronization can therefore overwrite previously valid snapshots and report success despite corrupted or attacker-modified manifests, undermining the integrity of the routing catalog and methodology summaries consumed by the research agent.

Content

No source excerpt is available for this finding.