Security audit

Claw Trace

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed tool-call tracing aid that is off by default, but users should enable detailed logs or file saving only when they are comfortable exposing trace data.

Install only if you want the agent to show its tool-call trace. Keep the default simple mode for normal work, avoid detailed logging around secrets or private files, review traces for accidental sensitive content, and enable workspace report saving only when you intentionally want a retained trace file.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Ssd 3

Medium

Confidence: 95% confidence
Finding: The skill requires automatic trace output after every tool call when enabled, which creates a continuous disclosure channel for tool inputs, outputs, errors, and operational context. Even with redaction guidance, automatic logging is dangerous because redaction is often incomplete and sensitive data can appear in unexpected fields or derived outputs.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The skill exposes broad natural-language commands such as 'enable XX feature' and similar phrases that can change runtime behavior without a tightly scoped control surface. In a tracing skill, unintended activation is risky because it can silently turn on logging or output features that expose tool inputs, outputs, or metadata during normal use.

Ssd 3

Medium

Confidence: 93% confidence
Finding: The detailed logging module instructs recording complete input/output for each tool call, which materially increases the likelihood of capturing secrets, personal data, internal prompts, or proprietary content. In the context of an agent tracing skill, this is especially dangerous because the skill's core purpose is to collect and present exactly the data that may contain sensitive material.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The skill documentation describes configuration options including detailed logging and saving reports to disk, but does not prominently warn users at the point of enablement that traces may include sensitive tool inputs/outputs and may be persisted. This increases the chance that users enable risky functionality without informed consent, leading to accidental disclosure or retention of confidential data.

Ssd 3

Medium

Confidence: 92% confidence
Finding: Saving trace reports to the workspace persists potentially sensitive tool-call contents beyond the immediate interaction, increasing exposure through later access, indexing, sharing, or backup. Persistence changes a transient disclosure risk into a retention and secondary-access risk, especially if reports contain detailed logs or partially redacted data.

Static analysis

No suspicious patterns detected.