Back to skill

Security audit

类型化知识图谱引擎

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed local knowledge-graph memory tool, but users should be aware it writes persistent local graph data and has an optional callback URL field with limited privacy detail.

Install this only if you want an agent to maintain a persistent local ontology under memory/ontology/. Avoid storing raw secrets in the graph, review repair or delete operations before allowing them, and do not provide a callback_url unless you trust the destination and understand what notification data your agent will send.

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
  • 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 (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill description includes a very broad activation phrase covering general AI models, agent orchestration, and LLM applications without narrow preconditions or clear trigger boundaries. In an agent environment, this can cause the skill to be invoked in situations where persistent writes, exec-based actions, or schema/file modifications were not intended, increasing the chance of unsafe side effects.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The input schema allows a callback_url for asynchronous notifications but does not warn that task data, graph contents, or metadata may be transmitted to an external endpoint. In agent workflows, this creates a data exfiltration and privacy risk because users may not understand that internal ontology contents could leave the local environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The markdown describes creating directories, writing append-only graph logs, validating schema, and even repairing corrupted files, but does not present a clear up-front warning that local persistent data and files will be modified. Users or higher-level agents may therefore invoke the skill assuming it is read-only, leading to unintended persistence, data corruption risk, or destructive repair actions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The display name, summaries, description, examples, and operational guidance are written almost entirely in Chinese, and there is no statement offering multilingual support or user language selection. This can constitute a language-policy issue when a skill effectively enforces one language without opt-in or justification.

Content

No source excerpt is available for this finding.