Back to skill

Security audit

Evolution档案 - 个人进化伴侣

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent local personal-growth tracker, but it persistently stores sensitive mood, habit, and reflection data in plaintext with broad triggers and weak reset safeguards.

Install only if you are comfortable with a local plaintext file retaining mood, habits, low points, and reflections over time. Before using it, consider tightening file permissions, backing up existing data, and avoiding broad or automatic recording unless you explicitly want that behavior.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:296
Finding

Plaintext Storage and Destructive Reinitialization of Sensitive Personal Records

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 208-209 and 296-325
Vulnerability Type: Plaintext sensitive-data storage, insufficient file-permission hardening, and unsafe file overwrite
Risk Level: Medium

The skill stores behavioral, emotional, habit, and personal-reflection data in a predictable plaintext JSON file:

markdown
### Storage Location
`~/.openclaw/data/evolution档案.json`

Initialization uses the following shell operation:

bash
mkdir -p ~/.openclaw/data
cat > ~/.openclaw/data/evolution档案.json << 'EOF'
{
  "initialized_at": "{{Current time}}",
  "last_update": "{{Current time}}",
  "profile": {
    "name": "大表哥",
    "streaks": {
      "早起": {"current": 0, "best": 0, "last_date": ""},
      "运动": {"current": 0, "best": 0, "last_date": ""},
      "阅读": {"current": 0, "best": 0, "last_date": ""},
      "写作": {"current": 0, "best": 0, "last_date": ""},
      "冥想": {"current": 0, "best": 0, "last_date": ""},
      "学习": {"current": 0, "best": 0, "last_date": ""},
      "复盘": {"current": 0, "best": 0, "last_date": ""}
    }
  },
  "heartflow": [],
  "troughs": [],
  "daily_mood": {},
  "achievements": [],
  "weekly_letters": [],
  "stats": {
    "total_heartflow_hours": 0,
    "total_troughs": 0,
    "current_best_streak": 0,
    "total_achievements": 0
  }
}
EOF

Technical Analysis

The archive may contain sensitive information such as mood scores, emotional low points, personal reflections, habits, achievements, and weekly letters. The initialization procedure does not establish restrictive permissions for either the directory or the JSON file. Actual permissions consequently depend on the process umask and preexisting filesystem state.

On systems with a permissive umask, other local accounts or processes may be able to read the archive. The predictable path also makes the data easy to locate. Plaintext storage further means ...[truncated 1816 chars]

Remediation
View remediation

Remediation Suggestions

  1. Create the data directory with owner-only permissions:

    bash
    install -d -m 0700 "$HOME/.openclaw/data"
    
  2. Set a restrictive umask before creating or updating the archive:

    bash
    umask 077
    
  3. Ensure the archive is owner-readable and owner-writable only:

    bash
    chmod 0600 "$HOME/.openclaw/data/evolution档案.json"
    
  4. Do not overwrite an existing archive silently. Test for its existence and require explicit user approval:

    bash
    archive="$HOME/.openclaw/data/evolution档案.json"
    if [ -e "$archive" ]; then
      printf '%s\n' "Archive already exists; initialization aborted." >&2
      exit 1
    fi
    
  5. Before any approved reset, create a timestamped backup with restrictive permissions and clearly inform the user that initialization will replace existing records.

  6. Write updates to a securely created temporary file in the same protected directory, validate the JSON, set mode 0600, and atomically rename it over the destination. This reduces corruption from interrupted writes.

  7. Consider encryption at rest for emotional and personal-reflection fields, with keys stored separately from the archive. Clearly document that local-only storage does not by itself guarantee confidentiality against other processes running with the user's privileges.

  8. Validate that ~/.openclaw/data and the archive are not symbolic links before writing, particularly if directory permissions may have been altered by another process.

Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (7)

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

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger phrases include very common expressions such as '今天怎么样' and '汇报', which can cause the skill to activate unintentionally during ordinary conversation. Because this skill persists sensitive mood, habit, and reflection data, accidental activation increases the chance of collecting and storing personal data without sufficiently explicit user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Module-level triggers like '状态很好', '太难了', '打卡', and '今天情绪' are ambiguous and can appear in normal chat unrelated to persistent tracking. In this skill's context, that ambiguity is more dangerous because the resulting actions involve storing longitudinal sensitive data across conversations.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill stores sensitive mood, habit, and reflective data locally, but the user-facing description and activation flow do not clearly warn that such data will be written to a persistent file. This creates a transparency and consent problem: users may disclose intimate information without understanding that it will be retained across sessions.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The defined JSON schema includes emotional states, low points, self-comfort text, achievements, and weekly reflective letters, all tied to timestamps and history. This is highly sensitive profile data, and the structured persistent archive makes it easy for anyone with local access to reconstruct intimate patterns, vulnerabilities, and routines.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
88% confidence
Finding

The initialization workflow explicitly creates a persistent file under ~/.openclaw/data and writes user profile data to it, establishing session persistence outside the active conversation. Persistence itself is not always unsafe, but here it stores sensitive personal content without any mention of access controls, permission hardening, encryption, or consent safeguards.

Content

Scanner excerpt · SKILL.md (reported line 296)May include surrounding context.

用户说"启动 evolution 档案",执行:

bash
mkdir -p ~/.openclaw/data
cat > ~/.openclaw/data/evolution档案.json << 'EOF'
{
  "initialized_at": "{{当前时间}}",

Ssd 3

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill is designed to continuously record and retain sensitive personal, emotional, and behavioral data over time, creating a detailed longitudinal profile of the user. Even if stored locally, such aggregation increases privacy risk, expands the consequences of local compromise, and may exceed what users expect from routine conversation.

Content

No source excerpt is available for this finding.