Back to skill

Security audit

Know My Larkmate

Security checks for vulnerabilities and agentic risk

Overview

The skill has a coherent Lark-to-notes purpose, but it asks for broad ongoing access to private workplace context and can automatically persist summaries into local memory notes.

Review before installing. Only use this if you are comfortable granting broad read-only Lark access, including messages, documents, meetings, wiki, calendar, and offline refresh access, and with recurring jobs copying selected work context into local OpenClaw memory notes. Prefer limiting enabled sources, avoiding unattended heartbeat use unless needed, pinning the CLI version, and revoking Lark authorization if you stop using the skill.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
scripts/request-readonly-auth.sh:4
Finding

Overbroad Persistent Lark Authorization and Automated Private-Context Collection

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:4
Finding

Unpinned Globally Installed Lark CLI Dependency

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (10)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The documented behavior focuses on scanning many Lark resources and validating tool readiness, but it does not define corresponding, constrained note-writing behavior in a way that can be audited against the declared purpose. In practice, this encourages broad data harvesting under the banner of 'context sync,' increasing privacy and scope-creep risk.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The documented behavior focuses on scanning many Lark resources and validating tool readiness, but it does not define corresponding, constrained note-writing behavior in a way that can be audited against the declared purpose. In practice, this encourages broad data harvesting under the banner of 'context sync,' increasing privacy and scope-creep risk.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
78% confidence
Finding

The skill instructs the agent to read local daily-note files (memory/YYYY-MM-DD.md, HEARTBEAT.md) and to use an external CLI, but it does not declare an explicit tool/permission boundary. That creates implicit file-read capability and weakens reviewability, making accidental overreach or unauthorized data access more likely in a memory-sync workflow handling potentially sensitive context.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation language is broad enough to match ordinary conversation, such as asking what the user has been doing recently. That can cause the agent to invoke external data reads and local note updates without a clear, specific request, leading to unintended access, retention, or summarization of sensitive work context.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Phrases like 'any normal turn that read relevant Lark artifacts' and 'update, refresh, or accumulate recent context' are ambiguous and operationally expansive. In a skill that accesses work communications and writes memory notes, such ambiguity materially increases the risk of unintentional triggering and silent persistence of sensitive information.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest advertises syncing recent Lark activity into OpenClaw daily notes but does not warn that the skill will import external context and modify persistent notes. This is risky because users may not realize that potentially sensitive workplace communications or summaries are being copied into another system, increasing the chance of unintended disclosure, over-collection, or persistence of private information.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill enables implicit invocation with only broad natural-language triggers in the metadata, which can cause the agent to run automatically in contexts the user did not explicitly intend. Because the skill imports external Lark context and writes into daily notes, unintended activation can silently copy sensitive work activity into notes or alter user data without clear consent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
80% confidence
Finding

The manifest describes syncing recent Lark context into daily notes, which aligns with appending relevant context. However, the heartbeat file also instructs broad retrospective maintenance actions like reviewing prior snapshots, merging duplicates, and pruning stale patterns, which goes beyond simple syncing recent Lark context and changes the behavior into ongoing note curation.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

This is a manifest-like file, so trigger quality applies. The file defines recurring tasks with generic names and free-form prompts, but does not document narrower activation constraints or exclusion conditions beyond time intervals, which can make task intent and invocation boundaries less explicit.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The prompts require the agent to reply with the fixed English token "HEARTBEAT_OK" when nothing changes. This imposes a specific language/output convention without offering user opt-in or documenting a justified locale constraint.

Content

No source excerpt is available for this finding.