Back to skill

Security audit

Clawddocs.Bak

Security checks for vulnerabilities and agentic risk

Overview

This is a Clawdbot documentation helper with mostly inert scripts and configuration examples, but users should review the network and token examples before copying them.

Before installing or using this skill, treat it as a lightweight documentation aid rather than a full working docs search system. Do not paste real bot tokens into shared files or logs, and avoid copying the 0.0.0.0 gateway example unless you intentionally want remote access and have authentication, TLS, and firewall controls in place.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
snippets/common-configs.md:37
Finding
Gateway Configuration Binds to All Network Interfaces Without Security Controls## Vulnerability Details **File Location**: `snippets/common-configs.md`, lines 37–45 **Vulnerability Type**: Insecure network exposure configuration **Risk Level**: Medium ### Vulnerable Code ```markdown ## Gateway Configuration ```json { "gateway": { "host": "0.0.0.0", "port": 8080 } } ``` ``` ### Technical Analysis The ready-to-use configuration binds the gateway to `0.0.0.0`, causing it to listen on every available IPv4 network interface rather than only the local loopback interface. The example does not include authentication, TLS, network allowlisting, or other access-control settings. This configuration alone does not prove that the gateway is unauthenticated or Internet-accessible, because actual exposure also depends on the gateway's defaults, host firewall, network topology, and perimeter controls. Nevertheless, users who copy the example may unintentionally make the service reachable from untrusted local networks or the Internet. ### Attack Path 1. A user copies the documented gateway configuration. 2. The gateway starts on TCP port `8080` and binds to all IPv4 interfaces. 3. The host firewall, cloud security group, container port mapping, or local network permits another system to reach that port. 4. An attacker discovers the exposed service through network scanning or service enumeration. 5. The attacker connects to gateway endpoints and attempts to use any functions not protected by effective authentication and authorization controls. 6. If such endpoints exist, the attacker may access gateway functions or sensitive bot-management capabilities available through them. ### Impact Assessment The maximum impact depends on the gateway implementation and its runtime security controls. Potential exposure includes unauthorized access to gateway APIs, administrative interfaces, bot-management operations, configuration data, or other functions available through the service. The con ...[truncated 311 chars]
Remediation
## Remediation Suggestions 1. Change the default example to bind only to the loopback interface: ```json { "gateway": { "host": "127.0.0.1", "port": 8080 } } ``` 2. Document `0.0.0.0` as an explicit remote-access option rather than the default. 3. Require strong authentication and authorization for every gateway endpoint before enabling remote access. 4. Protect remote traffic with TLS, preferably through a hardened reverse proxy or a documented native TLS configuration. 5. Restrict inbound connections using host firewalls, cloud security groups, container networking rules, or private-network allowlists. 6. Warn users not to expose port `8080` directly to the public Internet. 7. Add deployment guidance covering trusted proxies, rate limiting, request logging, credential rotation, and regular security updates.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The description claims a broad documentation assistant/tooling capability set, including navigation, search, fetching docs, version tracking, and config snippets. The supplied code chunk is much narrower: it merely echoes a URL based on the input path after checking that an argument was provided. This is a material mismatch because the declared primary purpose and capabilities substantially exceed what the code actually does, including even the claimed 'doc fetching' behavior.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The supplied code chunk is far narrower than the declared description. It only echoes a message about recently updated docs based on a days parameter and does not implement the broader claimed capabilities such as decision-tree navigation, search, documentation retrieval, version tracking, or configuration snippet generation. While comments mention change tracking, the actual code does not perform it. This is a material description-to-behavior mismatch because the declared primary purpose is much broader than the real behavior shown.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The activation scope is broad enough that the skill could trigger on ordinary Clawdbot-related conversation, causing the agent to adopt instructions that encourage shell-script use and document fetching without strong gating. In a skill system, overbroad activation increases attack surface by making it easier for untrusted user input to steer the agent into higher-risk tool usage or unexpected context switching.

Missing User Warnings

Low
Confidence
82% confidence
Finding
This markdown file includes example configuration for Discord and Telegram tokens, but it does not warn users that these values are sensitive credentials that must be stored securely and not committed or shared. Under the markdown-specific missing-warning rule, credential-affecting behavior should be accompanied by a privacy or security warning.

Static analysis

No suspicious patterns detected.