Back to skill

Security audit

1.1.0

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with medium confidence: the reported behavior appears aligned with a HiDataLinker database connector, though the install mechanism is noted and some references were not resolved.

The verdict is benign, but it is not a guarantee of safety. Before installing, consider that the instructions include building and copying a local agent binary, and that several referenced paths could not be resolved during the scan. Review the installation steps and decide whether you want to run them.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 19)May include surrounding context.

md
1. The HiDataLinker window must be open. If a tool says it is not running, ask the user to open the app. Do not try to connect with a host, user, or password yourself.
2. The MCP service must be on. In HiDataLinker open **Settings** and turn on **MCP service** / **MCP 服务**. The same switch is on the activity bar. If a tool returns `MCP service is turned off in HiDataLinker settings`, ask the user to turn that switch on. Turning it off rejects tool calls; it does not remove saved connections.
3. Use a saved `connectionId` from `list_connections`. Never ask the user for a database password. Never read `~/.hidatalinker/connections/` or `~/.hidatalinker/agent/endpoint.json`. The MCP process reads the endpoint itself.
4. If `connect` says the connection has no saved password, ask the user to connect that profile once in the desktop app. Do not request the password in chat.

## How to call

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 42)May include surrounding context.

sh
cargo build --bin agent-mcp --manifest-path src-tauri/Cargo.toml
mkdir -p ~/.hidatalinker/bin
cp src-tauri/target/debug/agent-mcp ~/.hidatalinker/bin/agent-mcp