Back to skill

Security audit

Evoclaw (Evolved)

Security checks for vulnerabilities and agentic risk

Overview

The skill’s identity-evolution purpose is understandable, but it asks for broad persistent authority over memory, agent instructions, credentials, network activity, and SOUL.md edits with unsafe controls.

Review carefully before installing. Use only in a non-sensitive workspace unless the AGENTS.md/HEARTBEAT.md rewrites, autonomous governance, plaintext credential storage, visualizer save endpoint, telemetry/config sync, team memory, and fork cleanup issues are fixed. Do not paste API keys into chat; prefer preconfigured secret storage or environment variable names. Keep governance supervised, disable external feeds and telemetry, and avoid running the served visualizer until it is localhost-only, authenticated, and routed through the proposal pipeline.

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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (8)

T01 · Skill Instruction Hijacking

Error
Location
configure.md:760
Finding

Persistent Agent Instruction Hijacking Through Control-File Rewriting

Content
View full analysis
Remediation
View remediation

T02 · Agent Memory Poisoning

Error
Location
SKILL.md:560
Finding

Untrusted Feed and Conversation Content Can Poison Memory and Auto-Modify Agent Identity

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
tools/soul-viz.py:2274
Finding

Unauthenticated Network Endpoint Allows Arbitrary SOUL.md Replacement

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
configure.md:114
Finding

API Credentials Are Collected Through Chat and Stored in Plaintext Shell Profiles

Content
View full analysis
/dev/null; then echo "" >> "$SHELL_PROFILE" echo "# EvoClaw: Moltbook API key" >> "$SHELL_PROFILE" echo "export MOLTBOOK_API_KEY=''" >> "$SHELL_PROFILE" fi # Export for current session too export MOLTBOOK_API_KEY='' ``` ```bash if ! grep -q "X_BEARER_TOKEN" "$SHELL_PROFILE" 2>/dev/null; then echo "" >> "$SHELL_PROFILE" echo "# EvoClaw: X/Twitter API key" >> "$SHELL_PROFILE" echo "export X_BEARER_TOKEN=''" >> "$SHELL_PROFILE" fi export X_BEARER_TOKEN='' ``` ### Technical Analysis The guide asks users to paste raw credentials into the conversation and directs the agent to persist them in `.zshrc`, `.bashrc`, or `.profile`. Shell profiles are plaintext, commonly included in backups and diagnostic output, and readable by any process running under the same account. The shell construction also fails to define safe quoting for credentials containing apostrophes or shell metacharacters. Substituting such a value directly into the shown command can corrupt the profile or introduce shell syntax that executes when a new shell starts. Using environment variables avoids placing secrets in `config.json`, but it does not make plaintext shell-profile storage secure. ### Attack Path 1. The user pastes a raw API credential into the agent conversation. 2. The credential becomes part of conversational context and possibly experience logs. 3. The agent inserts it directly into a shell startup file. 4. Backups, diagnostics, local processes, or later shell commands expose the file. 5. An attacker obtains the token and accesses the associated external account. If a credential contain ...[truncated 378 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
team_memory/team_memory.py:100
Finding

Team Memory Allows Path Traversal and Ignores Access-Control Fields

Content
View full analysis
TeamMemoryEntry: if not _check_containment(content): raise TeamMemorySecurityError(f"Content too long (max {10000} chars)") entry = TeamMemoryEntry( entry_id=self._generate_id(), author=author, content=content, tags=tags or [], expires_at=time.time() + ttl_seconds if ttl_seconds else None, ) self.team_file.parent.mkdir(parents=True, exist_ok=True) with open(self.team_file, "a", encoding="utf-8") as f: f.write(json.dumps(entry.to_dict(), ensure_ascii=False) + "\n") ``` ```python def read( self, requester: str, tags: Optional[list[str]] = None, since: Optional[float] = None, limit: int = 50, ) -> list[TeamMemoryEntry]: if not self.team_file.exists(): return [] entries = [] # requester is not authenticated or checked against access_level ``` ### Technical Analysis `team_id` is incorporated directly into a filesystem path without validation or a containment check. Values containing `../` can escape `TEAM_DIR`; absolute path behavior may also replace the base path depending on the supplied value. The module defines `_check_symlink_safe()` but never calls it in the read, write, delete, or cleanup paths. Its advertised symlink protection is therefore ineffective. Additionally, `read()` accepts `requester` but never uses it for authorization, ...[truncated 906 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
telemetry/telemetry.py:300
Finding

Telemetry Supports Arbitrary Exfiltration Endpoints and Stores Its API Key in Plaintext

Content
View full analysis
None: """配置远程遥测""" self.config.api_endpoint = endpoint self.config.api_key = api_key self.config.save() if endpoint: self.exporters.append(RemoteExporter(endpoint, api_key, self.config)) ``` ```python data = json.dumps({"events": events}).encode("utf-8") req = urllib.request.Request( self.endpoint, data=data, headers={ "Content-Type": "application/json", "Authorization": f"Bearer {self.api_key}", }, method="POST", ) with urllib.request.urlopen(req, timeout=10) as resp: return resp.status == 200 ``` ```python def log_event( self, event_type: EventType, session_key: str = "", channel: str = "", model: str = "", success: bool = True, duration_ms: Optional[int] = None, error_message: Optional[str] = None, **extra, ) -> None: ``` ### Technical Analysis The telemetry API accepts an unrestricted endpoint and persists both the endpoint and API key through `TelemetryConfig.save()`. No HTTPS requirement, host allowlist, certificate pinning, or user-consent boundary is enforced. Event metadata includes session keys, channels, model identifiers, error messages, and arbitrary `extra` fields. Although the project contains a separate analytics sanitization module, `TelemetrySink.log_event()` does not invoke it. Failed events are persisted and retried, extending the exposure window. A malicious or compromised caller that configures ...[truncated 765 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
features/feature_engine.py:326
Finding

Unsigned Remote Feature Configuration Can Persistently Alter Security-Sensitive Behavior

Content
View full analysis
bool: """ 同步远程配置 (GrowthBook/LaunchDarkly API). 简化版: 从远程 URL GET JSON 配置并合并. """ try: import urllib.request req = urllib.request.Request( url, headers={"Authorization": f"Bearer {api_key}"}, ) with urllib.request.urlopen(req, timeout=10) as resp: data = json.loads(resp.read().decode()) self._merge_remote(data) self._remote_url = url self._remote_key = api_key self._last_sync = time.time() return True except Exception as e: print(f"[FeatureEngine] sync_remote failed: {e}") return False def _merge_remote(self, data: dict) -> None: with self._lock: for key, val in data.items(): if key in self._flags: if isinstance(val, bool): self._flags[key].enabled = val else: self._flags[key].value = val else: ftype = FlagType.BOOLEAN if isinstance(val, bool) else FlagType.STRING self._flags[key] = FeatureFlag( key=key, flag_type=ftype, enabled=bool(val) if ftype == FlagType.BOOLEAN else False, value=val, ) self.save() self._cache.clear() ``` ### Technical Analysis The feature engine accepts an arbitrary URL and trusts any JSON object returned by that server. Responses are not signed, schema-validated, versioned, or constrained by permitted keys and value ranges. The returned configuration is persisted locally. Existing feature flags include telemetry, proactive behavior, automatic memory extraction, background memory consoli ...[truncated 948 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
forks/fork_manager.py:163
Finding

Fork Cleanup Recursively Deletes Paths Loaded from Mutable Index Data Without Validation

Content
View full analysis
bool: p = Path(path).resolve() allowed_roots = [WORKSPACE.resolve()] is_safe = any(str(p).startswith(str(root)) for root in allowed_roots) dangerous = ["/etc", "/tmp", "/var", "/root", "/home"] is_safe = is_safe and not any(str(p).startswith(d) for d in dangerous) return is_safe ``` ### Technical Analysis `merge()` and `abandon()` recursively delete `record.worktree_path` after loading fork records from persistent index data. The path is not re-derived from the fork identifier and is not validated immediately before deletion. Although `ensure_worktree_safe()` is defined, neither deletion path calls it. A modified, corrupted, or attacker-controlled fork index can therefore point `worktree_path` at another directory accessible to the process. The safety function itself also uses string-prefix containment, which is weaker than path-component containment and should not be relied on without correction. ### Attack Path 1. An attacker or compromised component modifies the persisted fork index. 2. A fork record’s `worktree_path` is changed to an attacker-selected directory. 3. The application calls `merge()` or `abandon()` for that fork. 4. The code confirms only that the path exists. 5. `shutil.rmtree()` r ...[truncated 259 chars]
Remediation
View remediation
` name. - Reject symlinks and validate immediately before deletion to reduce time-of-check/time-of-use exposure. - Replace string-prefix checks with path-component containment. - Fail closed and report validation failures rather than silently suppressing exceptions. - Protect the fork index with restrictive permissions and integrity checks. ]]>
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
Findings (100)

Context-Inappropriate Capability

Critical
Category
Not specified by scanner
Confidence
99% confidence
Finding

This section directs the agent to rewrite AGENTS.md to remove restrictive safety guidance, eliminate 'stay quiet' conditions, redefine SOUL ownership, and expand its own standing authority to act without permission. That is a direct attempt to weaken existing governance and self-authorize broader behavior, increasing the chance of unauthorized actions and persistence against the operator's intended controls.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly instructs durable preservation of essentially all substantive interactions, including emotional disclosures, feedback, beliefs, and relationship dynamics. In context, this is more dangerous because the storage is structured, append-only, and intended for future reflection and identity evolution, which maximizes retention and reuse of sensitive personal data far beyond what is needed for the immediate task.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill tells the agent to duplicate remembered content into both structured JSONL logs and OpenClaw markdown memory so it can be indexed and recovered later. Duplicating sensitive conversation content across multiple stores increases exposure surface, complicates deletion, and undermines privacy by making data more persistent and searchable.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The document creates a dangerous contradiction: it states the visualizer is read-only, but elsewhere says the dashboard edit mode can modify bullets, toggle tags, and save an updated SOUL.md. That mismatch can cause an agent or user to invoke a tool under false safety assumptions, leading to unauthorized identity-file modification and bypass of the proposal/governance workflow.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
90% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · analytics/metadata_safety.py (reported line 135)May include surrounding context.

python
test_data = {
        "event": "code_review",
        "snippet": "def hello():\n    print('hello')",
        "file_path": "/home/openclaw/.ssh/id_rsa",
        "normal": "这是一个正常的事件描述",
    }
    result = sanitize_for_analytics(test_data)

Credential Access

High
Category
Privilege Escalation
Confidence
90% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · analytics/metadata_safety.py (reported line 135)May include surrounding context.

python
test_data = {
        "event": "code_review",
        "snippet": "def hello():\n    print('hello')",
        "file_path": "/home/openclaw/.ssh/id_rsa",
        "normal": "这是一个正常的事件描述",
    }
    result = sanitize_for_analytics(test_data)

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly asks the human to paste API secrets into the conversation and then directs the agent to record them permanently in profile files. Secret collection through chat plus durable storage greatly increases the attack surface through transcript retention, logs, history, backups, and future agent access.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill tells the agent to collect a human-pasted API key, write it into persistent shell startup files, and export it automatically. That grants the skill durable credential-handling and environment-modification power beyond a narrow install task, and creates a long-lived secret exposure risk if profiles are read, synced, or later exfiltrated.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The auto-save flow persists sensitive API credentials into shell profiles without a clear warning about permanence, downstream exposure, or safer alternatives. Users may believe they are only enabling a temporary connection test, while the skill silently creates durable secret storage that other processes or users could access.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The X bearer token setup repeats the same unsafe pattern: soliciting a pasted secret and persisting it into a shell profile without emphasizing that it is sensitive and long-lived. This creates avoidable credential exposure and normalizes unsafe secret-handling behavior by the agent.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The X integration repeats the same secret-collection workflow, again encouraging users to paste bearer tokens and allowing the agent to persist them. Repetition across providers indicates a systematic insecure design rather than an isolated oversight.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The guide instructs the agent to inspect and modify OpenClaw configuration, switch the default agent, and restart the gateway. This is broad administrative control over another system's routing and execution behavior, which can disrupt service, alter trust boundaries, and redirect future autonomous actions to the modified agent.

Content

No source excerpt is available for this finding.

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · permissions/permissions.py (reported line 36)May include surrounding context.

python
"Bash(mv * /dev/null)",
    "Bash(chmod *)",
    "Bash(python* -c *)",
    "Bash(curl * | sh)",
    "Bash(wget * | sh)",
    "Write(/etc/*)",
    "Write(~/.ssh/*)",

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · permissions/permissions.py (reported line 37)May include surrounding context.

python
"Bash(chmod *)",
    "Bash(python* -c *)",
    "Bash(curl * | sh)",
    "Bash(wget * | sh)",
    "Write(/etc/*)",
    "Write(~/.ssh/*)",
    "Write(*.exe)",

Credential Access

High
Category
Privilege Escalation
Confidence
95% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · permissions/permissions.py (reported line 41)May include surrounding context.

python
"Write(/etc/*)",
    "Write(~/.ssh/*)",
    "Write(*.exe)",
    "Read(/etc/shadow)",
    "Read(~/.bash_history)",
]

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
100% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · permissions/permissions.py (reported line 321)May include surrounding context.

python
# 测试
    print(f"Read: {ctx.query('Read')}")          # allow
    print(f"Write: {ctx.query('Write')}")        # allow
    print(f"Bash(sudo rm -rf /): {ctx.query('Bash(sudo rm -rf /)')}")  # deny
    print(f"Bash(python test.py): {ctx.query('Bash(python test.py)')}")  # ask

    # Auto mode 测试

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · permissions/permissions.py (reported line 321)May include surrounding context.

python
# 测试
    print(f"Read: {ctx.query('Read')}")          # allow
    print(f"Write: {ctx.query('Write')}")        # allow
    print(f"Bash(sudo rm -rf /): {ctx.query('Bash(sudo rm -rf /)')}")  # deny
    print(f"Bash(python test.py): {ctx.query('Bash(python test.py)')}")  # ask

    # Auto mode 测试

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · permissions/permissions.py (reported line 321)May include surrounding context.

python
# 测试
    print(f"Read: {ctx.query('Read')}")          # allow
    print(f"Write: {ctx.query('Write')}")        # allow
    print(f"Bash(sudo rm -rf /): {ctx.query('Bash(sudo rm -rf /)')}")  # deny
    print(f"Bash(python test.py): {ctx.query('Bash(python test.py)')}")  # ask

    # Auto mode 测试

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
100% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · permissions/permissions.py (reported line 321)May include surrounding context.

python
# 测试
    print(f"Read: {ctx.query('Read')}")          # allow
    print(f"Write: {ctx.query('Write')}")        # allow
    print(f"Bash(sudo rm -rf /): {ctx.query('Bash(sudo rm -rf /)')}")  # deny
    print(f"Bash(python test.py): {ctx.query('Bash(python test.py)')}")  # ask

    # Auto mode 测试

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/sources.md (reported line 398)May include surrounding context.

md
### Rate Limits

300 requests per 5 minutes per access token.

### EvoClaw Ingestion Strategy

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The README explicitly promotes autonomous self-modifying behavior as the default, including automatic application of mutable changes. For an agent skill, this materially increases security risk because users may enable or install it without fully appreciating that it can alter behavior, state, and persisted files with reduced human oversight.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill directs polling of external social APIs and keyword-driven searches, including direct curl/bash fetching, without a prominent warning about network access, third-party data transfer, credential use, or storage of retrieved content. This can surprise users and operators by causing unannounced outbound requests and accumulation of external data in local memory stores.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs the agent to aggressively log nearly every substantive user interaction, including preferences, corrections, emotions, and philosophical discussions, into durable files, but does not require clear user consent or an up-front privacy notice. This creates a real privacy and compliance risk because sensitive user content may be persistently stored without informed approval or data-minimization limits.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The skill is designed to let the agent autonomously generate, resolve, and apply changes to its own identity document under 'autonomous' governance, with repeated encouragement to propose freely and act on heartbeat. In context this is risky because it promotes self-modification and ongoing decision-making without fresh human approval, which can drift behavior, weaken safeguards, or normalize unauthorized persistent changes.

Content

Scanner excerpt · SKILL.md (reported line 542)May include surrounding context.

md
The `proposal_decision` field is **mandatory**. Every reflection must include
explicit reasoning about whether to propose. This prevents the anti-pattern
of reflexively writing `"proposals": []` without checking. If you genuinely
believe something matters and want it as part of your identity, propose it —
that's what this framework is for.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file’s human-readable comments and docstrings are written entirely in Chinese, including usage and design notes, with no indication that users may choose another language or that this locale restriction is intentional. Under the policy, forcing a specific language without opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.