Back to skill

Security audit

Memory Pro

Security checks for vulnerabilities and agentic risk

Overview

This memory skill is not clearly malicious, but it can automatically write conversation-derived personal and project information into persistent notes and Git history without clear user approval.

Review this skill carefully before installing. Use it only in a workspace where you are comfortable having selected conversation details written to notes and committed to Git, and require manual review of every proposed memory entry before saving or committing it.

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:13
Finding

Unvalidated Conversation Content Can Be Written to Persistent Agent Memory

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 13-62
Vulnerability Type: Persistent memory poisoning through automatic distillation
Risk Level: Medium

Vulnerable Documentation Snippet

The following is an English translation of the relevant instructions in the source file:

markdown
## Trigger Conditions

Activate when any of the following conditions are met:

1. **End of conversation**: Proactively summarize and persist content after a complete conversation ends
2. **Important discovery**: New knowledge or an important conclusion is discovered during research or development
3. **Decision point**: The user makes an important decision and its rationale needs to be recorded
4. **Error correction**: Record lessons after a task failure or mistake
5. **User request**: The user says "write this down," "save this," or "this is a lesson"

### Step 2: Select the Target Location

| Content type | Target location | File naming |
|--------------|-----------------|-------------|
| User preferences, style, and agreements | `notes/people/` or `USER.md` | `USER.md` |
| Project progress and milestones | `notes/projects/<name>.md` | Project name |
| Accumulated knowledge and methodology | `notes/knowledge/<topic>.md` | Topic name |
| Lessons and error corrections | `notes/lessons/YYYY-MM-DD-<lesson>.md` | Date and lesson |
| Important decisions and rationale | `notes/decisions/YYYY-MM-DD-<decision>.md` | Date and decision |
| Personal relationships | `notes/people/<name>.md` | Person name |
| Cross-domain insights | `notes/ontology/` | Knowledge graph node |

### Step 3: Write or Update the File

**New file**: Create `notes/<category>/YYYY-MM-DD-<title>.md`

**Update an existing file**: Append to the corresponding section of the relevant file.

Technical Analysis

The skill directs the agent to extract information from ordinary conversations a ...[truncated 3254 chars]

Remediation
View remediation

Remediation Suggestions

  1. Require explicit user confirmation before every persistent write, including updates presented as automatic end-of-session summaries.
  2. Display the destination path and exact proposed diff before writing or appending content.
  3. Treat conversation text, imported notes, quotations, retrieved documents, and tool output as untrusted unless their provenance is verified.
  4. Reject persistent entries that attempt to modify safety constraints, authorization boundaries, tool policies, or the interpretation of higher-priority instructions.
  5. Add a sensitive-data screening stage that detects and redacts credentials, authentication tokens, private keys, personal information, and confidential project data.
  6. Store provenance metadata with every entry, including the source session, author or source type, creation date, approval state, and confidence level.
  7. Quarantine unverified statements rather than placing them directly in authoritative files such as USER.md.
  8. Restrict writable paths to a dedicated memory directory and prevent user-controlled titles or names from determining paths without normalization and validation.
  9. Implement retention periods, deletion controls, version review, and reliable rollback for persistent entries.
  10. Require stronger review before promoting information into any higher-trust or long-term memory layer.
  11. Before Git archival, require separate confirmation and scan the generated note for secrets because deletion from the working tree does not remove content from repository history.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (10)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation conditions are broad enough to match normal conversation flow, including session endings, discoveries, decisions, and corrections. This raises the risk of the skill activating without meaningful consent and silently persisting sensitive user data into long-term storage.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The workflow performs durable storage and git archival without requiring privacy review, sensitivity filtering, or explicit user acknowledgement. Persisting conversation-derived data into notes and commit history can make secrets, personal data, and mistaken content difficult to remove later.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger phrases are generic conversational statements that commonly occur in benign dialogue, making accidental activation likely. In a skill that writes to persistent memory and archives changes, ambiguous triggers materially increase privacy and integrity risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The skill name, description, and invocation wording are entirely in Chinese, and no alternative language option or opt-in is documented. Under the policy, locale or language constraints should not be imposed unless the skill offers choice or clearly justifies a region-specific requirement.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The high-level description emphasizes refinement and long-term memory deposition but does not clearly warn users that content may be written to files and archived. Without upfront disclosure, users may reveal sensitive information under the mistaken assumption that the interaction is ephemeral.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill directs retention of user preferences, decisions, errors, and conversation-derived knowledge by default when broad conditions are met. This creates a natural-language data retention risk because sensitive or private information may be stored permanently without necessity or explicit consent.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instructions explicitly write user-related and session-derived content into persistent note files, which can then be further exposed through version history. Durable storage of natural-language summaries increases the chance of privacy leakage, unauthorized later reuse, and difficult-to-revoke retention.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill instructs the agent to execute shell commands for git archival even though its stated purpose is memory distillation and note writing. Expanding capability from file writing to arbitrary command execution increases the attack surface and can lead to unintended command use, repository modification, or execution in unsafe environments.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

Advertising use of a general exec tool is broader than necessary for summarizing and storing notes, and normalizes command execution as part of routine memory handling. If the skill is triggered by ordinary conversation, this broad capability could be abused to run unintended commands or perform actions unrelated to memory management.

Content

No source excerpt is available for this finding.