Back to skill

Security audit

Conv Compactor

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only context compaction skill with no code execution or data exfiltration behavior, though its trigger wording could cause unwanted summarization.

Install only if you want the agent to summarize and compress long conversations. Prefer invoking it explicitly, and review generated summaries for missing details before relying on them for ongoing work.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manual trigger phrases include very common conversational terms such as “总结一下”, which can naturally appear in ordinary dialogue without an actual intent to invoke compaction. In this skill, unintended activation can compress or replace live context at the wrong time, causing loss of fidelity, omission of active work, or propagation of an incomplete summary into subsequent turns.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The condition “对话明显变慢或开始重复之前的内容” is subjective and not operationally defined, so different agents or implementations may activate the skill inconsistently. In a context-compaction skill, ambiguous activation criteria increase the chance of premature summarization, which can degrade reliability and accidentally discard important conversational state.

Content

No source excerpt is available for this finding.