Back to skill

Security audit

Bluecollar-Hr Officers

Security checks for vulnerabilities and agentic risk

Overview

This HR toolkit needs review because its workflows require retaining identity-document photos and sharing arrival photos without clearly defining their limits.

Review the photography steps before using real staff or student data. Prefer minimal identity-verification records; where photos are necessary, define consent, authorized recipients, secure storage, and deletion dates, with appropriate safeguards for minors. The supplied static scan reported no findings; this Review verdict rests on the documented data-handling instructions.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

The transfer workflow requires identity checks and photography without a privacy warning.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The transfer workflow mandates identity verification and photography (“核身份证拍照”) without specifying whether identity documents or passengers are photographed, whether image collection is necessary, or how images must be protected and deleted. In a personnel-transfer workflow potentially involving students and minors, implementing this requirement could cause unnecessary collection and retention of sensitive identity information, increasing exposure to unauthorized disclosure or identity misuse. Identity verification has a legitimate operational purpose, and the reference to general compliance guidance may provide additional safeguards, but those safeguards are not established in the supplied file.

Content

No source excerpt is available for this finding.

The onboarding workflow instructs taking photos and posting them to a group without a privacy warning.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The onboarding checklist explicitly directs staff to take photos and post them to a group without specifying authorized recipients, necessity, consent or another applicable legal basis, or retention controls. In a workflow involving students, including minors, identifiable photos can expose attendance, location, and employment associations to unintended recipients; the file's general confidentiality-labeling requirement does not adequately control this sharing. The instruction does not explicitly require photographing identity documents, but its proximity to ID verification also creates ambiguity about what may be photographed.

Content

No source excerpt is available for this finding.

Mandatory arrival photography lacks a privacy notice.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The arrival procedure mandates roll call and photography without specifying notice, a lawful basis, access restrictions, or retention limits for identifiable photographs. In a labor-placement context that may involve students and minors, these images could link individuals to travel and employment details, creating privacy risks if retained or shared improperly; protections elsewhere for printed personal-information forms do not explicitly cover these photographs.

Content

No source excerpt is available for this finding.

The routing table uses everyday phrases and generic keywords without an explicit domain-context gate.

Medium
Category
Trigger Abuse
Confidence
91% confidence
Finding

L211 routes phrases such as “帮我安排” (“help me arrange”) and generic terms such as “客户” (“customer”) to X0; other rows similarly use broad words such as “网络” (“network”), “数据” (“data”), and “计划” (“plan”). Although the catalog concerns staffing operations, the routing instructions do not explicitly require that context or exclude unrelated requests, leaving these terms liable to cause unintended routing.

Content

No source excerpt is available for this finding.

The checklist instructs users to photograph and retain identity documents without a privacy warning or collection notice.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

该条目要求对每人的身份证原件拍照留档,却未说明完整影像收集的必要性、处理依据、告知程序或保存期限,容易将身份核验扩大为不必要的敏感个人信息留存。蓝领用工及学生实习场景涉及批量人员和未成年人,会放大泄露、身份冒用及违规处理的风险;后文的加密存储与外发审批要求虽有缓解作用,但不能替代收集最小化和明确的处理规则。

Content

No source excerpt is available for this finding.

The arrival workflow instructs staff to photograph participants and share the photos in a group without a privacy warning.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The arrival checklist explicitly requires photographing participants and posting the images to a group, but specifies no necessity assessment, notice, lawful basis, recipient restrictions, or retention limits. In a labor-placement workflow potentially involving students and minors, identifiable arrival photos could disclose participants’ location and employment or school affiliations to unnecessary recipients and be copied or forwarded; the internal-use label does not establish access controls.

Content

No source excerpt is available for this finding.

The trigger list includes generic framework terms without requiring a blue-collar HR context.

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

The trigger list includes generic terms such as “能力单元,” “身份叠加,” and “meta-skill-system” without requiring blue-collar HR context, potentially causing this skill to activate for unrelated requests. Unintended activation could impose its HR-specific routing, mandatory reference loading, and output conventions on those requests; however, the supplied file shows no malicious payload or sensitive-data transfer, limiting the demonstrated impact to task interference.

Content

No source excerpt is available for this finding.