Back to skill

Security audit

Cognition Evolution

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malicious code, but it automatically turns conversations into long-term memory without clear user approval, scoping, or deletion controls.

Install only if you are comfortable with automatic long-term memory creation from conversations. Before enabling it, prefer a version that requires explicit confirmation, excludes sensitive data, records source/provenance, and lets you review or delete stored memories.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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
Findings (1)

T02 · Agent Memory Poisoning

Warning
Location
SKILL.md:3
Finding

Automatic Deposition of Untrusted Conversation Content into Long-Term Memory

Content
View full analysis
Remediation
View remediation
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 (9)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The listing describes automatic end-of-conversation review and cross-session accumulation of lessons, but it does not clearly warn users that conversation-derived content may be persisted beyond the current session. This creates a meaningful privacy and consent risk because users may disclose sensitive personal, business, or team information without realizing it will be retained and reused later.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The listing explicitly promises automatic end-of-conversation review, lesson extraction, and cross-session retention, but it does not warn users that conversation-derived content may be persisted over time. This creates a privacy and consent risk because users may share sensitive business, personal, or operational details without realizing they will be retained and reused later.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill markets persistent retention of lessons, decisions, and unresolved issues across sessions, which can easily include sensitive user-provided content. Without clear scope limits, minimization, or handling rules, stored summaries may expose confidential information later through reuse, leakage, or unintended recall in future conversations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The entire user-facing documentation and manual trigger phrases are presented only in Chinese, implying a fixed language/locale expectation without any opt-in or alternative. Under the policy, language constraints should be optional or clearly justified for a region-specific tool.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README states the skill is 'fully automatic' and triggers after every conversation end, but it does not define scope boundaries, consent, exclusions, or content sensitivity rules. For a memory-writing skill, this can cause over-collection and persistence of sensitive or irrelevant conversation data, increasing privacy and prompt-scope risks.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The listed trigger phrases ('复盘', '总结', '认知升级') are common conversational terms and can appear in ordinary discussion without the user intending to activate persistent memory behavior. Since this skill stores derived insights to long-term memory, accidental activation creates a real risk of unauthorized or unexpected retention of conversation content.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill auto-triggers at 'session end' including heartbeat, which is an overly broad activation condition for a write-capable behavior. Because the skill performs persistence ('写入长期记忆'), ambiguous automatic invocation can cause unintended memory writes, capture low-quality or sensitive content, and make it hard for users to control when summarization/persistence occurs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The natural-language content of the skill file is fully Chinese and does not state that the user can choose another language or that the skill is specifically limited to Chinese-speaking users. Under the language/locale policy, forcing a specific language without opt-in can be a policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The entire skill description and required output format are specified only in Chinese, with no indication that the user can choose another language or that the skill is intended solely for a Chinese-language environment. This creates a natural-language locale constraint without explicit opt-in or justification.

Content

No source excerpt is available for this finding.