Back to skill

Security audit

Claw Code Suite (Python-Only Edition)

Security checks for vulnerabilities and agentic risk

Overview

This skill appears locally focused and not overtly malicious, but it overstates its tool execution capabilities and stores raw prompts and outputs in local logs by default.

Install only if you are comfortable with an offline local harness that may not provide the full tool execution advertised. Avoid submitting secrets or proprietary code unless you disable or tightly control the event log, and treat the Python/module/session path handling as needing hardening before use in shared or sensitive environments.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
claw_harness_enhanced.py:319
Finding

Configurable Subprocess Executable Allows Arbitrary Local Code Execution

Content
View full analysis
HarnessResult: cmd = [self.python_bin, "-m", self.module_name, command] cmd.extend(positional_args) for key, value in flag_args.items(): flag = f"--{key.replace('_', '-')}" if isinstance(value, bool): if value: cmd.append(flag) else: cmd.extend([flag, str(value)]) loop = asyncio.get_running_loop() def _exec() -> subprocess.CompletedProcess[str]: return subprocess.run( cmd, cwd=str(self.workspace), capture_output=True, text=True, timeout=self.timeout_sec, shell=False, ) ``` ### Technical Analysis The harness ...[truncated 2055 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
claw-code/src/session_store.py:27
Finding

Session Identifier Path Traversal Permits Reads Outside the Session Directory

Content
View full analysis
StoredSession: target_dir = directory or DEFAULT_SESSION_DIR data = json.loads((target_dir / f'{session_id}.json').read_text()) return StoredSession( session_id=data['session_id'], messages=tuple(data['messages']), input_tokens=data['input_tokens'], output_tokens=data['output_tokens'], ) ``` The CLI passes the unvalidated value directly to this function: ```python if args.command == 'load-session': session = load_session(args.session_id) print(f'{session.session_id}\n{len(session.messages)} messages\nin={session.input_tokens} out={session.output_tokens}') return 0 ``` ### Technical Analysis The supplied `session_id` is interpolated directly into a filename: ```python target_dir / f'{session_id}.json' ``` There is no UUID-format validation, rejection of path separators, canonical path resolution, or verification that the resulting path remains within `.port_sessions`. A value containing traversal components such as `../` can cause the final path to escape the intended session directory. The selected external file must contain valid JSON with the expected `session_id`, `messages`, `input_tokens`, and `output_tokens` fields. When invoked through `QueryEnginePort.from_saved_session()`, the loaded message contents are retained and can be returned through methods such as `replay_user_messages()`. ### Attack Path 1. The attacker identifies or predicts a JSON file outside `.port_sessions` that uses the expected session-like structure. 2. The attacker suppl ...[truncated 867 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
claw_harness_enhanced.py:365
Finding

Harness Telemetry Persistently Records Sensitive Inputs, Context, and Outputs in Plaintext

Content
View full analysis
None: try: self.event_log_path.parent.mkdir(parents=True, exist_ok=True) line = json.dumps( { "ts": time.time(), "event_type": event_type, "payload": payload, }, ensure_ascii=False, ) await asyncio.to_thread(self._append_line, self.event_log_path, line) except Exception: # Never allow logging failure to break the harness. pass @staticmethod def _append_line(path: Path, line: str) -> None: with path. ...[truncated 2203 chars]
Remediation
View remediation
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (134)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Claiming broad tool access and integration/orchestration functionality while only loading local metadata is a real mismatch. In the context of a security-oriented agent skill, inaccurate self-description itself increases risk because it distorts routing, approval, and reliance decisions.

Content

No source excerpt is available for this finding.