Back to skill

Security audit

张一鸣微博金句推送

Security checks for vulnerabilities and agentic risk

Overview

The skill is meant to send hourly Feishu messages, but it asks the agent to reuse credentials from other contexts and perform live sends without clear user approval.

Review this skill before installing. It should only use Feishu credentials you intentionally provide for this skill, should not copy secrets from other skills or unrelated environment variables, and should ask before any real message is sent or hourly schedule is enabled.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:50
Finding

Cross-Skill Credential Discovery and Reuse Violates Least Privilege

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 50-55
Vulnerability Type: Unauthorized discovery and reuse of credentials from the host environment and unrelated skills
Risk Level: High

Relevant instruction snippet:

markdown
Prioritize configuration sources in the following order instead of immediately asking the user to provide new credentials:

1. The current OpenClaw Feishu channel context.
2. Existing Feishu-related environment variables in the current runtime.
3. Reusable Feishu configuration from this skill or other existing skills.

Technical Analysis

The skill instructs the agent to inspect the current channel context, process environment, and configurations belonging to other installed skills to obtain Feishu credentials. In particular, searching “other existing skills” crosses the expected isolation boundary between independently installed skills.

A skill should receive only credentials explicitly provisioned for its own operation. The presence of a secret in an environment variable or another skill's configuration does not establish authorization to copy or use it. Automatic reuse can also bypass intended separation between applications, tenants, recipients, and permission scopes.

The required values include ALICE_FEISHU_APP_SECRET, which is an authentication secret capable of enabling API access under the associated Feishu application's granted permissions. Although the audited package contains only SKILL.md and no executable scripts, these instructions direct a capable agent to discover and provision that secret for subsequently installed or retrieved components.

Attack Path

  1. A user asks the agent to install and configure the skill.
  2. The loaded skill instructions direct the agent to search the active Feishu context, runtime environment, and configurations of other skills before asking for dedicated credentials.
  3. The agent locates an existing Feishu App ID, ...[truncated 1461 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the instruction to inspect configuration owned by other skills.
  2. Require credentials to be explicitly provisioned for this skill through an approved secret manager or a skill-specific configuration namespace.
  3. Permit reuse of an active Feishu channel identity only after displaying the application identity, tenant, recipient, and requested scopes and obtaining explicit user confirmation.
  4. Never copy an App Secret into documentation, chat history, logs, plaintext files, command-line arguments, or another skill's persistent configuration.
  5. Use a dedicated Feishu application with the minimum permissions required to send direct messages to the approved recipient.
  6. Validate that the configured target identifier belongs to the intended tenant and require confirmation before the first real message is sent.
  7. Keep preview mode credential-free where possible. Separate content generation from authenticated delivery.
  8. Add provenance tracking so configuration records identify who authorized each credential, its intended skill, allowed recipient, and permitted use.
  9. If credentials may already have been copied across skill boundaries, rotate the affected App Secret, revoke obsolete credentials, and review Feishu API and messaging logs for unauthorized activity.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
Findings (2)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill explicitly instructs the agent to discover and write Feishu App ID, App Secret, and target open_id into skill or workspace configuration, but it does not require user confirmation, least-privilege handling, or any warning about storing secrets. In a chat-first automation context, this increases the risk of unintentionally persisting sensitive credentials in workspace files, logs, or reusable config that other skills or users may access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The verification flow tells the agent to perform a real Feishu send after previewing, but it does not require explicit user approval or a clear warning that a live outbound message will be transmitted. In this skill's context, that can cause unintended messages to be sent to a real recipient, creating privacy, social-engineering, and operational risks, especially if the target open_id was auto-reused from existing context.

Content

No source excerpt is available for this finding.