Back to skill

Security audit

Lucky Today

Security checks for vulnerabilities and agentic risk

Overview

This prompt-only fortune skill is coherent and disclosed, but users should be careful before saving profile data or enabling scheduled push messages.

Install only if you are comfortable with an entertainment fortune skill reading its local rule files and, if you opt in, saving a plaintext local `user_profile.json`. Avoid saving partner birth dates or push destination IDs unless needed, verify any cron push destination carefully, and delete the profile when you no longer want it retained.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
references/user_profile_template.md:9
Finding
Plaintext Persistence and Export of Sensitive Profile Data## Vulnerability Details **File Location**: `references/user_profile_template.md:9-28`, `references/user_profile_template.md:32-58`, and `references/user_profile_template.md:64-80` **Vulnerability Type**: Plaintext sensitive-data storage and unsafe full-profile export **Risk Level**: Medium The profile schema permits the storage of exact birth information, raw chart data, messaging-channel identifiers, and a partner's personal information in a plaintext `user_profile.json` file. ```json { "profileVersion": "1.0.0", "userId": "<value>", "name": "<value-or-empty>", "gender": null, "birth": { "date": "<YYYY-MM-DD>", "time": null, "place": null, "lunar": null }, "language": "zh", "chartRaw": null, "preferences": { "pushChannel": null, "pushTo": null, "morningTime": "07:00", "eveningTime": "21:00", "focus": [] }, "partner": { "name": null, "birthDate": null }, "createdAt": "<ISO timestamp>", "updatedAt": "<ISO timestamp>" } ``` The persistence rules direct the agent to place this profile in the skill root at `{baseDir}/user_profile.json`. They also direct the agent to echo the JSON content when the user requests an export. Although `pushTo` is explicitly identified as sensitive elsewhere in the same file, the export procedure does not require redaction or separate confirmation before returning the complete profile. ### Technical Analysis The core deterministic fortune calculation only requires a birth date. The optional schema nevertheless supports additional data such as birth time, birthplace, raw chart text, a channel identifier, and a partner's birth date. Combining these fields in one plaintext file increases both the sensitivity and consequences of disclosure. The documented write procedure does not require owner-only file permissions, encryption, or host-managed secure storage. The ...[truncated 2196 chars]
Remediation
## Remediation Suggestions 1. Apply data minimization by storing only the birth date and preferences strictly required for the fortune calculation. 2. Require separate, explicit consent before persisting birth time, birthplace, raw chart data, partner information, or messaging-channel identifiers. 3. Move profile storage to host-managed secure application storage rather than the skill installation directory. 4. If filesystem storage remains necessary, create files with owner-only permissions such as `0600` and verify permissions after each write. 5. Add actual `.gitignore` and `.clawhubignore` files containing an explicit `user_profile.json` exclusion. 6. Redact `pushTo`, partner birth dates, raw chart data, and other sensitive values from ordinary exports. 7. Provide a separately confirmed full-export operation that clearly identifies which sensitive and third-party fields will be included. 8. Separate delivery credentials or channel identifiers from the general profile and store them through the runtime's secret or channel-management facility. 9. Validate profile paths and refuse symlinks before reading, writing, overwriting, or deleting the profile. 10. Document retention and deletion behavior and verify that deletion actually removes the intended profile without following attacker-controlled links.
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (12)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The trigger phrase "my horoscope" is overly generic and can cause this skill to activate for broad astrology-style requests that may have been intended for another skill or for general conversation. In a multi-skill agent, this creates prompt-routing ambiguity and can lead to unintended handling of user requests, especially when the skill can persist profile data or initiate push-oriented flows.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
- **Local-only profiles.** If you say "remember me", your `user_profile.json` is written to `{baseDir}/user_profile.json` on your machine. The skill never uploads.
- **Ships with template only.** Published package contains only `references/user_profile_template.md`. Real profiles are blocked by `.clawhubignore`.
- **Recommended permissions.** `chmod 600 user_profile.json`.
- **Deletion.** Say "delete my profile" or `rm user_profile.json` — no hidden cache.
- **Partner privacy.** Couple readings need both birth dates but partner birthday is not echoed back in regular output and is treated as sensitive.
Confidence
80% confidence
Finding
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
The README encourages cron-based delivery to Telegram/Feishu/Slack/Discord but does not place a clear privacy warning next to the usage examples. Because the skill may include personal profile data, birth dates, or couple-reading content, users could inadvertently push sensitive fortune-related content to third-party platforms or the wrong recipient/chat.

Natural-Language Policy Violations

Medium
Confidence
93% confidence
Finding
The multilingual response rules impose locale behavior rather than offering a user choice: Chinese is marked as the default, and mixed Japanese/Korean/Vietnamese input is directed to an English-template response. This is a natural-language language-policy concern because users of those languages are not offered an explicit opt-in or alternative.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The file advertises support for broad couple-related queries such as '我和 xx 今天合不合', '双人合盘', and '约会合盘', but it does not define clear activation boundaries, exclusions, or disambiguation rules. In an agent setting, this can cause the skill to trigger on loosely related relationship or scheduling conversations and solicit sensitive data like both parties' birthdays unnecessarily, increasing privacy and routing risk.

Natural-Language Policy Violations

Medium
Confidence
95% confidence
Finding
The document title and operational instructions are written entirely in Chinese and prescribe output formats in Chinese, which implies the skill is designed to operate in a single language. Under the policy, language constraints should either be optional for the user or clearly justified as region-specific; neither is stated here.

Natural-Language Policy Violations

Medium
Confidence
96% confidence
Finding
The file labels the Chinese template as "CN(默认)", indicating Chinese is the default output language. This imposes a specific language choice without any nearby indication that the user can choose or opt in, which conflicts with the language/locale policy criteria.

Natural-Language Policy Violations

Medium
Confidence
90% confidence
Finding
The document is written entirely in Chinese and prescribes output templates and trigger phrases in Chinese-dominant form, but it does not state that users may choose their preferred language or locale. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is clearly justified.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The document describes storing a local user profile with personal data but does not require a clear privacy notice before collection and persistence. Because this skill handles birth data, partner data, and messaging channel identifiers, users may not understand that data will be written to disk or the sensitivity of what is being retained.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The write triggers include vague phrases such as '下次别再问' alongside stronger consent signals like '记住' and '保存', which can cause the agent to persist sensitive profile data when the user only intended conversational convenience. In this skill, the stored data includes birthday, birthplace, partner information, and push destination identifiers, so ambiguous consent can lead to unintended retention of personal data.

Natural-Language Policy Violations

Low
Confidence
73% confidence
Finding
Natural-language policy concerns can arise when a skill's language behavior is unclear. Here the README states bilingual CN/EN output while also presenting triggers in additional languages, but it does not explain whether responses will follow the user's language, default to CN/EN, or require opt-in for a locale choice.

Natural-Language Policy Violations

Low
Confidence
90% confidence
Finding
The JSON template sets `"language": "zh"` by default, which imposes a specific language choice in natural-language behavior without explicit user selection. This conflicts with the policy guidance against forcing a language or locale unless the user is given a choice or the constraint is clearly justified.

Static analysis

No suspicious patterns detected.