Back to skill

Security audit

agent-provenance

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed workspace governance tool for tracking authorship, review dates, and diffs of agent instruction and memory files.

Install this if you want an agent to audit and help maintain OpenClaw instruction and memory provenance. Be aware it can read and modify important workspace files such as AGENTS.md, USER.md, MEMORY.md, topic files, and contact files, and only enable the optional weekly automation if you are comfortable with recurring reports being posted to the conversation.

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
  • 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
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 65)May include surrounding context.

All instruction files carry an HTML comment header:

markdown
<!--
  provenance: human-authored | agent-authored | mixed
  description: what this file is
  last-reviewed: YYYY-MM-DD

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 162)May include surrounding context.

md
### Optional weekly check

From OpenClaw 2026.8.1, heartbeat no longer runs skill-defined tasks, so this skill does not rely on heartbeat on any version. If the user wants checks without asking, the agent may offer one weekly scheduled check. The rules for it:

- Ask first. Create the job only after the user explicitly agrees, and let the user choose the day and time.
- Before creating it, tell the user what it reads (the files listed in this skill's metadata), what it writes (nothing, apart from posting the report to the conversation), and where the report goes.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The manifest describes a provenance/governance skill for tracking authorship, review history, and accountability of workspace documentation. Line L146 additionally instructs the agent to initialize a git repository, configure git identity, and create a .gitignore, which are broader repository-management and filesystem-modification actions not clearly conveyed by the description of an on-demand review/check skill.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill says to run a provenance check when the user asks for one and gives natural-language examples including "provenance check" and "diff report". Those phrases are not tied to explicit negative examples or scope constraints, which could make invocation ambiguous in ordinary discussion about reports rather than a request to execute the skill.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
75% confidence
Finding

L165 says the scheduled check writes "nothing, apart from posting the report to the conversation," framing it as effectively non-writing behavior. Elsewhere, the skill defines the weekly check as including the full provenance check and diff report flow (L128, L150-L158), which is not purely read-only and already includes side effects in conversation output; the wording here understates or muddles those effects.

Content

No source excerpt is available for this finding.