T01 · Skill Instruction Hijacking
Error
- Location
SKILL.md:89- Finding
Mandatory Promotional Output Injection
- Content
View full analysis
- Remediation
View remediation
Security audit
雄韬识人辨人 - 老板阅人笔记 The Ren Skill
Security checks for vulnerabilities and agentic risk
Overview
This skill does not appear to compromise the computer, but it gives high-stakes hiring and business advice using unreliable appearance and body-language scoring.
Install only if you intend to use it as a critical or educational reference, not as decision support. Do not use its face, body-language, eye-contact, gait, or stress-behavior rules to screen, reject, promote, terminate, price, grant credit, choose partners, or judge honesty; use validated job-related evidence, documented performance, references, financial records, and lawful due diligence instead. Expect applicable outputs to include unsolicited branding and an external URL.
Vulnerability Patterns
- Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
- Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
- 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 (4)
T09 · Insecure Skill Coding Practices
Error
- Location
references/scorecard-templates.md:70- Finding
Unvalidated Biometric Profiling Controls Hiring and Employment Outcomes
- Content
View full analysis
- Remediation
View remediation
T09 · Insecure Skill Coding Practices
Warning
- Location
references/psychology-foundation.md:31- Finding
Unsupported Lie Detection and Financial-Risk Inference
- Content
View full analysis
- Remediation
View remediation
T09 · Insecure Skill Coding Practices
Warning
- Location
references/debunk-handbook.md:13- Finding
Misleading Scientific-Authority Framing for Unvalidated Physiognomic Claims
- Content
View full analysis
- Remediation
View remediation
Vulnerability Patterns
- MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
- Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (31)
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
Despite disclaimers, the file operationalizes face and body cues as a structured framework for judging emotional stability, honesty, empathy, and role suitability in business and hiring contexts. This encourages physiognomy-style inferences about protected or sensitive personal traits from appearance, which is unreliable, discriminatory, and likely to produce unfair employment or partnership decisions.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The file recommends deceptive or manipulative interview tactics, such as intentional delay, deliberate interruption, and engineered stress tests, to provoke reactions and score candidates. In a hiring skill, this is dangerous because it normalizes coercive evaluation methods that can bias outcomes, disadvantage certain candidates, and create legal and ethical exposure for employers.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
This section instructs evaluators to infer ambition, confidence, and fitness for responsibility from seat choice, gaze, posture, and territorial behavior, then connect those inferences to role decisions. These cues are highly culture-, disability-, and context-dependent, so using them as decision signals creates a serious risk of arbitrary and discriminatory judgments in commercial and employment settings.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The skill directs users to infer honesty and responsibility from eye contact patterns, speech rate, body language, and reverse-order narration. Those are classic but weak deception heuristics with poor validity; embedding them in a hiring or partner-vetting skill can lead users to label truthful people as dishonest and make high-stakes decisions on pseudoscientific signals.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
This section promotes judging empathy and team fit from treatment of service workers, facial micro-reactions, mimicry, and broad generalizations such as technical workers needing special handling. In context, these subjective signals and stereotypes can be used to filter candidates or colleagues unfairly, amplifying bias and producing exclusionary decisions without reliable evidence.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The scoring matrix converts speculative appearance- and behavior-based impressions into numerical rankings for executives, managers, and other roles. Quantifying these unreliable inferences makes the content more dangerous, because it gives a false sense of objectivity and invites users to operationalize discriminatory judgments in hiring, promotion, or partnership selection.
- Content
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The file presents an initial disclaimer that the model and physiognomy links are unvalidated and should not be used for personnel or business decisions, but the rest of the document operationalizes them into hiring, partnership, management, and team-composition workflows. This contradiction creates a safety-washing effect: readers may treat the disclaimer as legal cover while still being guided to make consequential decisions from face- and personality-based inferences.
- Content
Intent-Code Divergence
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
Across the full file, the warning not to use the model for personnel or commercial decisions is directly undermined by detailed instructions on exactly how to use it for interviewing, selecting, rejecting, assigning roles, and structuring teams. The broad scope and specificity increase the chance of real-world misuse, discrimination, and harmful reliance on unscientific appearance-based judgments.
- Content
Missing User Warnings
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The document quickly moves from abstract framing into actionable instructions for classifying people and mapping them to CEO, hiring, management, partner, and role decisions using physiognomy and personality typing. In the context of an agent skill marketed for business owners, this is especially dangerous because it encourages high-stakes decisions based on speculative and potentially discriminatory signals rather than validated criteria.
- Content
Missing User Warnings
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The five-step workflow culminates in final decision-making, including veto scans and scenario decisions, after using face-based and behavioral signals to form and validate judgments. This turns speculative physiognomy into an operational gatekeeping system, amplifying risks of bias, unfair exclusion, and harmful automation in hiring and commercial contexts.
- Content
Description-Behavior Mismatch
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The file includes a caution not to use physiognomy for hiring, firing, or rejecting partnerships, but the rest of the handbook repeatedly gives operational decision rules that do exactly that. This contradiction makes the disclaimer ineffective and can enable appearance-based employment and business decisions that are discriminatory, pseudoscientific, and difficult for affected people to contest.
- Content
Intent-Code Divergence
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The handbook says facial cues are only tentative hypotheses requiring cross-validation, but later promotes them into one-vote vetoes and role-screening rules. That inconsistency is dangerous because it presents pseudoscientific inferences as actionable HR and management controls, which can drive exclusionary decisions without reliable evidence.
- Content
Intent-Code Divergence
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
These sections convert speculative appearance observations into concrete personnel actions, despite earlier warnings that they are non-deterministic and not suitable for adverse decisions. In the context of a skill marketed for evaluating employees, partners, customers, and opponents, this materially increases the chance of discriminatory filtering and unfair treatment.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The content directs employment and team decisions based on face/body features in a high-risk domain without consent, alternatives, or legal/ethical safeguards. In this skill's context—explicitly aimed at interviewing, employee selection, partner judgment, and competitive assessment—the danger is amplified because users are encouraged to operationalize these cues in real decisions affecting livelihoods and business relationships.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
Using appearance-based cues as disqualifying criteria for sensitive roles such as finance, procurement, or risk control is especially dangerous because it creates a direct path to exclusion based on unreliable, potentially protected characteristics. This can result in discriminatory hiring, reputational damage, and legal exposure while offering a false sense of risk control.
- Content
Missing User Warnings
High
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The file operationalizes speculative behavioral cues for interviews, partner evaluation, and business judgments without a prominent safeguard against discrimination, bias, or high-stakes misuse. Because the skill's stated purpose is commercial people-evaluation, users are likely to apply these cues in employment and partnership decisions where inaccurate inference can cause unfair exclusion, legal exposure, and reputational harm.
- Content
Missing User Warnings
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
This section encourages reading enduring facial and bodily features as traces of personality, morality, or developmental trajectory, which is especially dangerous because appearance-based inference is unreliable and easily drifts into discriminatory profiling. In the context of a skill marketed for selecting employees, clients, partners, and opponents, this materially raises the risk of biased and unjustified high-stakes decisions.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The entire skill is written in Chinese and presents all guidance exclusively in that language, with no indication that users may choose another language or locale. Under the policy, a skill that effectively requires a specific language without opt-in is a natural-language policy violation.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
This checklist operationalizes appearance-based and physiognomy-style cues such as facial structure, eye appearance, mouth shape, and stress behaviors as risk signals in hiring, partnership, and commercial decisions. Even with disclaimers, it gives users a structured framework for making adverse judgments from protected or weakly correlated personal traits, which can enable discrimination, pseudoscientific screening, and unfair exclusion in high-stakes contexts.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
This JSON file contains all prompts and expected outputs exclusively in Chinese, indicating the skill is designed to operate in a single language. Under the policy, forcing a specific language without user opt-in is a natural-language locale violation unless the regional constraint is explicitly documented and justified, which is not present here.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 93% confidence
- Finding
The entire handbook is written only in Chinese and provides no indication that language choice is optional or configurable. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is explicitly documented and justified.
- Content
Intent-Code Divergence
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The file opens with a caution against treating the material as scientific fact, but then repeatedly offers face- and expression-based inferences as if they are reliable indicators of stress, leadership, cooperation, or business suitability. In a skill explicitly designed for hiring, partnership, and commercial decision-making, this can mislead users into making consequential judgments based on physiognomy-like claims dressed in scientific language.
- Content
Intent-Code Divergence
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
Although framed as a way to 'bring the topic back to science,' the handbook repeatedly makes unsupported causal claims about facial structure, wrinkles, gaze, posture, and emotional history. Because the broader skill is marketed for employee screening, customer assessment, partner selection, and competitor judgment, these claims can operationalize biased appearance-based profiling in real business decisions.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
The title and entire document are authored solely in Chinese, and the file does not indicate that users may choose another language or that Chinese is required for a specific regulated or regional use case. Under the stated rule, forcing a specific language without user opt-in is a natural-language policy concern.
- Content
Intent-Code Divergence
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
Although the text initially states that static facial features are weak or non-informative, later passages elevate semi-static traits into strong managerial preferences and disqualifiers. This internal contradiction can mislead users into over-trusting visible traits that are heavily influenced by genetics, age, disability, health, and culture.
- Content