Back to skill

Security audit

Compact Test A

Security checks for vulnerabilities and agentic risk

Overview

The skill is purpose-aligned, but it saves broad conversation and tool-output details into durable memory without strong privacy, consent, or redaction rules.

Install only if you are comfortable with the agent saving compacted details from your conversations and tool results into local memory files. Review or edit the skill first to require explicit approval before memory writes and to forbid saving tokens, passwords, cookies, private keys, authorization headers, full connection strings, personal data, or raw sensitive tool output.

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:34
Finding

Potential Plaintext Persistence of Sensitive Conversation and Tool-Output Data

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The document labels the usage and explanation section as "中文说明" and presents the skill instructions primarily in Chinese, while also defining Chinese trigger phrases as first-class activation terms. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy concern unless the locale restriction is clearly justified or optionality is offered.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 49)May include surrounding context.

方式三:只复制核心文件

bash
mkdir -p ~/.openclaw/skills/smart-compact
curl -o ~/.openclaw/skills/smart-compact/SKILL.md \
  https://raw.githubusercontent.com/wavmson/openclaw-skill-smart-compact/main/SKILL.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 50)May include surrounding context.

bash
mkdir -p ~/.openclaw/skills/smart-compact
curl -o ~/.openclaw/skills/smart-compact/SKILL.md \
  https://raw.githubusercontent.com/wavmson/openclaw-skill-smart-compact/main/SKILL.md

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill defines proactive and automatic activation conditions such as suggesting itself when context nears 80%, which can cause the workflow to run without a clear, user-initiated need. In this skill, unintended invocation matters because later phases direct the agent to scan prior tool outputs and persist extracted information, increasing the chance of unnecessary data collection and retention.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The principle '先救再压' and the wording '宁可多存也不能漏存' create a default bias toward over-retention of information from the conversation and tool outputs. In a compaction skill, this is particularly risky because the material being 'rescued' may include secrets, internal paths, credentials, user data, or other sensitive operational details that should expire rather than be preserved.

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 write extracted information from tool outputs and conversation into memory files, but it does not require filtering for secrets, personal data, credentials, tokens, or system prompts before persistence. Because memory files are durable, this can turn transient sensitive context into long-lived stored data that may later be surfaced or reused inappropriately.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

These instructions direct collection and persistence of broad categories including addresses, configuration values, endpoints, file paths, user preferences, and task progress. Those categories often contain sensitive or security-relevant data, and the skill provides no guardrails for consent, minimization, redaction, or classification before writing them to durable memory.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The rule stating that key information must never be omitted encourages the agent to prioritize completeness over privacy and security. In context, that makes the skill more dangerous because it is specifically designed to inspect large tool outputs before compaction, so the policy incentivizes persisting exactly the kinds of transient sensitive details that should often be discarded.

Content

No source excerpt is available for this finding.