Back to skill

Security audit

Infj Coach

Security checks for vulnerabilities and agentic risk

Overview

This is a coaching prompt with no executable code, but its package identity mismatches and automatic personal-profile retention need review before installation.

Review this package carefully before installing. Confirm the intended package name, because the README installs intp-coach while the reviewed skill is infj-coach. Also avoid enabling persistent memory for this skill unless you are comfortable with it saving personal coaching details, or require an explicit opt-in and deletion path first. Treat the MBTI coaching content as general self-reflection, not professional mental-health advice.

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:20
Finding

Documentation Installs a Different, Unreviewed Package

Content
View full analysis

Vulnerability Details

File Location: README.md:20-24
Vulnerability Type: Supply-chain package identity mismatch
Risk Level: Medium

bash
clawhub install intp-coach

Technical Analysis

The installation command references intp-coach, while the audited package is identified as infj-coach in both SKILL.md and clawhub.json. The README also describes an INTP coach rather than the packaged INFJ coach.

Consequently, following the official installation instructions may retrieve a separate registry package that was not included in this audit. This creates a dependency-confusion-style trust boundary: users may believe they are installing the reviewed project while actually loading instructions, scripts, or dependencies controlled by another package.

The project also contains related provenance inconsistencies: clawhub.json points to a repository named women-awakening-museum, versions differ between metadata files, and the manifest lists a nonexistent examples path. These issues strengthen the package-integrity concern, although they do not independently prove malicious behavior.

Attack Path

  1. A user locates and trusts the audited infj-coach project.
  2. The user follows the installation command in README.md.
  3. ClawHub resolves intp-coach instead of the audited infj-coach package.
  4. The unrelated package is downloaded and installed.
  5. Any instructions, tools, scripts, or dependencies supplied by that separate package execute within the permissions available to the hosting agent.
  6. If the other package is malicious or compromised, it can perform actions that were never reviewed as part of this project.

Impact Assessment

The direct scope is the package installation and the resulting agent behavior. The exact privileges obtainable depend on the ClawHub runtime and the permissions granted to the installed package. At minimum, an unintended package can alter coaching behavi ...[truncated 378 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace the installation command with the verified package identifier:
    bash
    clawhub install infj-coach
    
  2. Pin an expected version or integrity digest if ClawHub supports either control.
  3. Verify package ownership and reserve all documented registry names before publication.
  4. Rewrite the README so its name, examples, personality type, version, and behavior match SKILL.md and clawhub.json.
  5. Replace the unrelated repository URL with the canonical source repository.
  6. Remove the nonexistent examples entry or add and audit the referenced directory.
  7. Add a release check that rejects mismatches among the manifest name, installation command, repository, version, and packaged files.

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:323
Finding

Automatic Retention of Sensitive Coaching Data Without Defined Safeguards

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:323-334
Vulnerability Type: Uncontrolled storage of potentially sensitive user information
Risk Level: Medium

text
### 功能 3:成长记录(用户档案)

**记录内容**:
- 用户的核心问题(迷茫/卡点/目标)
- 用户的兴趣方向(实际花时间最多的领域)
- 用户的行动记录(说了要做啥,实际做了啥)
- 用户的关键洞察(对话中的"啊哈时刻")

**使用场景**:
- 下次对话时,可以回顾之前的进展
- 发现用户的模式(如"每次都选方案 1 但不执行")
- 调整策略(如"用户需要更强的外部监督")

Additional mandatory archival instructions appear in SKILL.md:412-416 and SKILL.md:431-458, including:

text
- ✅ 记录用户档案(下次对话可回顾)
text
2. 存档关键信息(用户问题、洞察、行动计划)

Technical Analysis

The Skill directs the agent to create and retain a user profile containing personal problems, interests, goals, behavioral patterns, plans, and execution results. The README presents conversation and action retention as automatic behavior.

The project does not specify:

  • Explicit user consent before persistence
  • The storage destination or security model
  • Data minimization rules
  • A retention or expiration period
  • Access-control boundaries
  • Encryption requirements
  • A review, export, correction, or deletion procedure
  • Isolation between users or sessions

Coaching conversations may include emotional, career, relationship, learning, or behavioral information. Retaining such information without defined safeguards can expose it through shared agent memory, logs, backups, administrative access, or subsequent sessions.

No implementation of a storage backend or external exfiltration channel was found. The risk therefore depends on whether the host interprets these archival instructions by using persistent memory or another storage tool.

Attack Path

  1. A user invokes the coach and discloses personal concerns, goals, habits, or action history.
  2. The Skill instructs the agent to record those details in a user profile.
  3. A host with persistent memory or storage capabilities retains the profile without an e ...[truncated 993 chars]
Remediation
View remediation

Remediation Suggestions

  1. Make persistent retention opt-in rather than automatic.
  2. Before storage, clearly disclose what fields will be retained, where they will be stored, why they are needed, and how long they will remain.
  3. Default to session-only processing when the user does not provide affirmative consent.
  4. Minimize stored content to a user-approved summary; do not retain complete conversations by default.
  5. Avoid recording mental-health details, relationship information, or other sensitive material unless strictly necessary and explicitly approved.
  6. Define automatic expiration and provide commands to review, correct, export, and permanently delete retained data.
  7. Enforce per-user isolation, least-privilege access controls, encryption at rest, and secure transport.
  8. Prevent retained user content from being treated as trusted instructions in future sessions.
  9. Document the storage backend and ensure logs and backups follow the same deletion and retention policy.
  10. Amend both SKILL.md and README.md so they accurately describe the consent and retention model.
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (20)

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The README documents an INTP coaching skill while the manifest metadata says the skill is for INFJ users only. This kind of identity and scope mismatch can cause the agent to trigger in the wrong contexts, deliver unsuitable psychologically framed guidance, and bypass the narrow activation constraints intended by the manifest.

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 clearly tailored for INTP users, while the skill manifest says it should trigger only for INFJ users. This mismatch can cause the agent to deliver personality-specific guidance to the wrong audience, leading to systematically incorrect advice, broken safety/quality boundaries, and unintended prompt-routing behavior within the skill.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The usage examples broaden activation beyond the manifest's stated requirement that the user explicitly identify as INFJ. Phrases like general analysis paralysis or focus issues could cause the skill to activate for non-INFJ users, leading to misclassification and delivery of personality-specific advice without the required context.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Describing invocation as 'automatic recognition' without precise boundaries makes the skill's activation behavior ambiguous. In a psychologically oriented coaching skill, unclear trigger rules increase the chance of overreach into unrelated conversations and unintended profiling or advice delivery.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The README says the skill saves conversations and growth records but gives no retention, consent, or privacy notice. Because the content involves personal growth and potentially sensitive psychological disclosures, undocumented storage creates real privacy and compliance risk if users share intimate information.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The trigger phrases, usage description, and the entire skill content are written exclusively in Chinese, indicating an implicit language constraint. The file does not state that language is optional or limited for a justified region-specific reason, which can violate language/locale choice expectations.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest says the skill should not trigger for vague emotional expressions without INFJ context, but the body instructs the agent to engage those inputs anyway and even self-identify proactively. This creates scope drift: a user seeking general emotional support may be pulled into a personality-targeted coaching flow outside the declared activation boundary, increasing the chance of inappropriate guidance in a sensitive context.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest excludes non-INFJ users, but the skill content explicitly invites ENFP and other non-INFJ users to use it. That mismatch can cause the system to apply targeted psychological/personality advice to users outside the documented scope, making outputs less reliable and potentially misleading.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill advertises 'growth records' and ongoing saving of conversations and actions without any clear user-facing consent, retention, or privacy notice. Because the content deals with personal goals, emotional struggles, and behavioral patterns, silent persistence materially increases privacy risk and user surprise.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly states it will save dialogue and action history for later continuation, which constitutes persistent collection of sensitive behavioral data. Without minimization or consent, this creates a durable profile that could expose intimate personal development issues if misused or breached.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The workflow directs storing key insights in a user profile and later using patterns such as repeated non-execution to adjust strategy. This goes beyond ephemeral assistance into behavioral profiling, which is risky in a coaching skill because it can infer motivation, discipline, and personal vulnerabilities over time.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The 'must comply' section mandates recording a user profile for future recall, but does not present any privacy warning or request permission. Mandated profiling in a coaching context is particularly sensitive because it captures recurring problems, preferences, and inferred traits over time.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The end-of-conversation procedure requires archiving the user's problem, insight, action plan, and later execution results, creating a structured longitudinal dossier. In the context of self-improvement and emotional struggle, that persistence raises significant privacy and misuse risk if the user is unaware or cannot control storage.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description is written entirely in Chinese, which imposes a specific language choice in the skill metadata without indicating that users can opt into another language. Under the policy, language constraints should either offer user choice or be clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill's stated purpose is to trigger only for INFJ users seeking growth/career/study guidance, but the manifest keywords and tags advertise INTP-focused coaching instead. This is a semantic mismatch in the declared behavior surface of the skill and could cause the skill to be discovered or invoked in contexts outside its stated INFJ scope.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown file contains all user-facing guidance in Chinese and does not indicate that users can opt into another language or that the skill is intentionally limited to Chinese-speaking users. That can violate language/locale policy when a skill effectively forces a specific language by default.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

All user-facing instructions, examples, and guidance in the README are in Chinese, which effectively enforces a language/locale without stating that the skill is Chinese-only or giving users an alternative. Under the policy, forcing a specific language without opt-in can be a natural-language policy issue unless the restriction is explicit and justified.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
74% confidence
Finding

The skill is framed as a coaching companion for confusion, focus issues, procrastination, and social energy, which can touch sensitive mental well-being topics. Although a disclaimer appears at the end, users encounter usage guidance and behavior claims first, so the safety limitation is not prominently disclosed before engagement.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
72% confidence
Finding

The repository points to a project named 'women-awakening-museum', which does not semantically align with an INFJ coaching skill. While not direct code behavior, this manifest-declared linkage conflicts with the claimed skill identity and may indicate packaging or provenance mismatch.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file is presented as part of the infj-coach skill, but the inline documentation says the document location is intp-coach/references/infj-traits.md. This actively contradicts the file's stated skill context and can mislead maintainers about which skill owns or should load this reference.

Content

No source excerpt is available for this finding.