Back to skill

Security audit

Esfj Coach

Security checks for vulnerabilities and agentic risk

Overview

This looks like a coaching prompt rather than malware, but its package identity is inconsistent and it tells agents to save sensitive coaching profiles without clear consent controls.

Review before installing. The package should correct its name, README, install command, repository URL, tags, and reference files so they consistently describe the same ESFJ skill. It should also make saved coaching records opt-in, show users what will be saved, and provide clear review/delete controls.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
README.md:18
Finding

Mismatched Installation Command Can Install an Unreviewed Package

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:283
Finding

Automatic Persistence of Potentially Sensitive Coaching Profiles Without Defined Safeguards

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (17)

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The README identifies the skill as an INTP coach while the manifest says the skill is for ESFJ users, creating a material mismatch between declared purpose and documented behavior. This can cause the wrong skill to be invoked or trusted for the wrong audience, leading to unsafe or inappropriate guidance and undermining user and platform expectations about what the skill does.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Nearly the entire documented behavior targets INTP-specific coaching patterns rather than the manifest's ESFJ-focused purpose, indicating a broad scope integrity failure rather than a minor typo. In an agent ecosystem, this can misroute sensitive self-help interactions, produce systematically mismatched advice, and defeat any policy review or user consent based on the manifest description.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

Marking 'growth record' as 'automatic execution' implies stateful storage or logging may occur without a clearly bounded user action. In a coaching context, this is especially risky because conversations may contain sensitive emotional, behavioral, or mental-health-adjacent information that should not be retained by default.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The README states that conversations and actions are saved but provides no clear warning about data retention, privacy consequences, or handling of sensitive personal content. This creates a meaningful privacy risk because users may disclose intimate self-improvement or psychological concerns without understanding that the information is being persisted.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest metadata materially conflicts with the stated purpose of the skill: the package is named 'esfj-coach' but its description, keywords, and tags point to an INTP-focused coaching tool. This kind of identity mismatch is dangerous because users, reviewers, or automated systems may trust the declared skill purpose while actually installing or invoking something different, which increases the risk of deceptive packaging, incorrect routing, and hidden malicious or inappropriate behavior in associated files.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The reference file is entirely focused on INTP coaching, while the skill metadata says the skill is for ESFJ-specific growth guidance. This mismatch can cause the agent to deliver guidance optimized for the wrong personality type, undermining reliability and potentially causing harmful or misleading advice in a coaching context where users may rely on tailored psychological framing.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

All user-facing instructions, examples, and behavior descriptions are written in Chinese, and the README does not indicate that users can choose another language. This can constitute a language-policy issue when a skill effectively forces a specific language without opt-in or justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Describing mode selection as 'automatic recognition' without clear trigger boundaries or exclusions leaves activation behavior ambiguous. Ambiguous invocation rules can cause the skill to engage unexpectedly, choose the wrong mode for sensitive conversations, or interfere with other skills because users and reviewers cannot predict when it will act.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill advertises a "成长记录" capability that saves dialogue and action history, which implies persistence of potentially sensitive personal data such as emotional state, goals, and behavioral patterns. Because the skill does not clearly warn the user up front, obtain consent, or define storage scope/retention, users may disclose sensitive information without understanding it will be archived.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This section instructs the agent to record a user profile containing core problems, interests, action records, and key insights from coaching conversations. In a coaching context, these details can reveal sensitive psychological and personal information, and collecting them without explicit notice and consent increases privacy risk and the chance of inappropriate profiling or retention.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The end-of-conversation workflow directs the system to archive key user information automatically at session close, including the user's problem, insights, action plan, and outcomes. Automatic archival at the end of a vulnerable coaching exchange is risky because users may interpret the interaction as ephemeral unless clearly told otherwise and given a chance to decline.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The repository URL points to an unrelated project ('women-awakening-museum') rather than a repository that clearly corresponds to the declared ESFJ coaching skill. An unrelated source reference breaks provenance and traceability, making it harder to verify authorship, review source history, or detect substituted content; in a supply-chain context, this can conceal tampering or facilitate trust abuse.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
79% confidence
Finding

This markdown reference is written entirely in Chinese and is described as internal guidance for the skill at runtime. Because the skill consumes it as behavioral context and there is no indication that users can opt into another language or that the skill is intentionally region-specific, this may impose a language/locale constraint without user choice.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The document is labeled as ESFJ guidance, but the traits and function descriptions largely map to a different profile, creating a semantic integrity failure in the skill’s reference data. In a coaching skill, this can systematically produce inaccurate or psychologically mismatched advice, especially because the file is explicitly meant for internal guidance rather than direct user scrutiny.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown file contains end-user skill reference content exclusively in Chinese, and nowhere indicates that the user can choose another language or that the skill is intentionally region- or locale-specific. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file explicitly labels itself as 'INTP Coach' reference data in the later sections, directly contradicting the ESFJ intent declared in the manifest. This strengthens evidence that the skill may be wired to the wrong behavioral framework, increasing the chance of systematic misclassification and inappropriate user guidance.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The documented file path points to a different coach directory, which indicates copy/paste contamination and weak configuration integrity. While this is less severe than the profile mismatch, it increases the chance of loading or maintaining the wrong reference material and signals that the skill may be wired to incorrect assets.

Content

No source excerpt is available for this finding.