Back to skill

Security audit

Workflow Crystallizer

Security checks for vulnerabilities and agentic risk

Overview

This skill is coherent and locally focused, but it persistently caches memory-log content and turns untrusted memory headings into agent-readable automation drafts without enough safeguards.

Install only if you are comfortable with this skill reading your OpenClaw memory logs and keeping a local plaintext cache of memory-derived excerpts. Review state.json periodically, avoid running it automatically until its triggers are narrowed, and treat generated cron definitions and skill drafts as untrusted drafts that need human review before approval.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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 (2)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/analyze_patterns.py:232
Finding

Plaintext Persistent Caching of Sensitive Memory Content

Content
View full analysis
list[dict]: """Parse a memory file into structured events (one per H2 section).""" text = filepath.read_text(encoding="utf-8", errors="replace") ``` ```python return { "section": header, "keywords": keywords, "actions": actions, "entities": extract_entities(full_text), "has_steps": detect_steps(body), "time_hint": extract_time_hint(header, body), "day_of_week": day_of_week, "is_formalized": detect_formalized(header, body), "raw_summary": body[:500].strip(), } ``` ```python events = parse_memory_file(filepath) add_events(state, date_str, events) new_event_count += len(events) ``` ```python def save_state(state: dict, path: Optional[str] = None) -> None: """Write state to disk atomically.""" p = Path(path) if path else DEFAULT_STATE_PATH p.parent.mkdir(parents=True, exist_ok=True) tmp = p.with_suffix(".tmp") with open(tmp, "w") as f: json.dump(state, f, indent=2, default=str) tmp.replace(p) ``` ### Technical Analysis The analyzer reads Agent memory files and stores up to 500 characters from each parsed section in the `raw_summary` field. These excerpts are added to the persistent `event_cache` and serialized into `state.json` as plaintext. Atomic replacement protects the state file from partial writes but does not protect confidentiality. The implementation has no secret redaction, encryption, data minimization, expiration policy, or explicit restrictive file permissions. The resulting file permissions depend on the process umask and execution environment. Because Agent memory may contain private conversations, project information, credentials, tokens, inte ...[truncated 1361 chars]
Remediation
View remediation

T01 · Skill Instruction Hijacking

Warning
Location
scripts/analyze_patterns.py:97
Finding

Untrusted Memory Headings Propagated into Agent-Readable Reports and Skill Drafts

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Memory Manipulation

High
Category
Memory Poisoning
Confidence
90% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · scripts/state.py (reported line 54)May include surrounding context.

python
state.setdefault("analysis_log", [])
            return state
        except (json.JSONDecodeError, KeyError):
            sys.stderr.write(f"Warning: Corrupt state file at {p}, starting fresh.\n")
            return EMPTY_STATE.copy()
    return EMPTY_STATE.copy()

Memory Manipulation

High
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · scripts/state.py (reported line 248)May include surrounding context.

python
import argparse
    parser = argparse.ArgumentParser(description="Inspect crystallizer state")
    parser.add_argument("--state-file", default=None, help="Path to state.json")
    parser.add_argument("--reset", action="store_true", help="Reset state to empty")
    args = parser.parse_args()

    if args.reset:

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill documents and encourages scripts that read memory logs and write persistent state, but it declares no explicit tool scope or permissions boundary. That creates a mismatch between described capabilities and declared access, increasing the chance the agent is invoked with broader filesystem access than intended and enabling unintended reads or writes to sensitive local data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad and overlap with common requests like workflow optimization, habits, recurring tasks, and automation suggestions, making accidental or overly frequent invocation more likely. Because this skill persists state and analyzes memory logs, overbroad triggering can cause it to process sensitive historical data in contexts where the user may not have intended that level of analysis.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This markdown template instructs generated skills to use a one-line description 'with trigger words' and a 'Trigger phrases: [keywords from cluster]' section without requiring specificity, constraints, or negative examples. That can lead to skill activations based on broad keyword clusters that overlap with normal conversation, which matches the vague-trigger category for markdown files.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
80% confidence
Finding

The skill persists analysis history, event cache, and suggestions to disk in a predictable JSON file without any permission hardening or confidentiality controls. In this skill context, that state likely contains sensitive user workflow and memory-log derived data, so other local users or processes could read or tamper with it if filesystem permissions are weak.

Content

Scanner excerpt · scripts/state.py (reported line 60)May include surrounding context.

python
def save_state(state: dict, path: Optional[str] = None) -> None:
    """Write state to disk atomically."""
    p = Path(path) if path else DEFAULT_STATE_PATH
    p.parent.mkdir(parents=True, exist_ok=True)
    tmp = p.with_suffix(".tmp")

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The footer string forces the report timestamp to be labeled as 'ET', which imposes a specific locale/timezone in user-facing output. The file does not offer a timezone choice or explain why Eastern Time is required, which fits the natural-language locale policy violation criteria.

Content

No source excerpt is available for this finding.