Back to skill

Security audit

Daedalus

Security checks for vulnerabilities and agentic risk

Overview

ClawScan’s verdict is benign with medium confidence: the recorded findings are treated as expected for the skill’s purpose, though Skillspector reported undeclared capabilities and its review was partial.

The verdict is benign, but review the skill’s stated purpose against the reported ingestion and installation capabilities, and consider whether its broad activation and undeclared tool scope fit your needs. Skillspector’s analysis is partial: four local path-like references were unresolved, so the evidence is incomplete.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (8)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code's primary function is directory-content aggregation, not test-first implementation, debugging, validation, or diff review. Although aggregation could assist code investigation, the declared description does not disclose this bulk-ingestion function or establish its role in the stated workflow. The code performs local filesystem reads and writes; it shows no network transmission or execution of ingested content.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code's primary function is repository ingestion, which the description does not disclose. It performs no testing, code modification, incremental validation, or diff review. Although ingestion could assist a later investigation, this standalone bulk cloning and export capability is materially different from the declared method. The code shows no upload of extracted content or execution of cloned source files.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The supplied command performs skill installation rather than the declared coding workflow. It contains no test-first implementation, debugging, measurement, or diff-review behavior. Although installation could support later use of the described skill, it is the entire behavior shown and modifies the agent's skill environment—an installation capability not disclosed in the description. The behavior of the downloaded skill cannot be verified from this snippet.

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/ingest_github_repo.py (reported line 61)May include surrounding context.

python
try:
            # Using subprocess.run for better control and error handling
            process = subprocess.run(git_command, capture_output=True, text=True, check=False) # check=False to inspect output

            if process.returncode == 0:
                console.print(f"[green]Successfully cloned repository to temporary location: {clone_dir}[/]")

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 · tests/test_tools.py (reported line 28)May include surrounding context.

python
(self.source / "ignored.bin").write_bytes(b"\x00\xff")

    def run_tool(self, name, *args):
        result = subprocess.run(
            [sys.executable, str(SCRIPTS / name), *map(str, args)],
            cwd=self.root, capture_output=True, text=True, timeout=30,
        )

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 · tests/test_tools.py (reported line 61)May include surrounding context.

python
def test_repository_ingestion_clones_selected_branch(self):
        def git(*args):
            subprocess.run(["git", "-C", str(self.source), *args], check=True,
                           capture_output=True, text=True)

        git("init", "-b", "fixture")

Implicit activation depends on a subjective condition that can encompass nearly any coding task.

Low
Category
Trigger Abuse
Confidence
91% confidence
Finding

The skill activates when a code change or diagnosis 'benefits from a failing test or falsifiable measurement first,' which does not distinguish eligible tasks from ordinary coding requests. Unlike explicit invocation by name, this alternative provides no clear boundary for when the method should not activate.

Content

No source excerpt is available for this finding.