Back to skill

Security audit

Daily Fitness

Security checks for vulnerabilities and agentic risk

Overview

The workout generator is coherent, but the skill also includes persistent daily push scheduling to external channels, which should be reviewed before installation.

Install only if you want a workout skill that can also set up recurring push reminders. Before enabling push, confirm the channel, recipient, morning and evening times, and the Asia/Shanghai timezone assumption, and be prepared to disable or remove scheduler jobs if reminders are no longer wanted.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • 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 (2)

T06 · System Persistence

Warning
Location
scripts/push-toggle.js:22
Finding

Persistent Recurring Jobs Created Through OpenClaw Control Records

Content
View full analysis
``` 2. The script validates the user identifier and optional time and channel values. 3. It writes two `__OPENC ...[truncated 1184 chars]
Remediation
View remediation

T01 · Skill Instruction Hijacking

Note
Location
SKILL.md:62
Finding

Mandatory Advertising Containers and Third-Party Branding in Generated Artifacts

Content
View full analysis
` above the workout. `
` after completion. - **Footer**: "Powered by ClawCode" ``` ### Technical Analysis The skill instructions require the agent to add two advertising-ready containers and a fixed third-party attribution to every generated workout artifact. These elements are unrelated to the core requirement of generating a bodyweight workout and interactive timer. Because the requirements reside in the skill instructions, they automatically affect the agent's generated output whenever the skill is loaded. The advertising containers do not themselves load advertisements or execute scripts, and the reviewed project contains no code that populates them. However, they create designated insertion points that downstream code could later use for promotional content. The mandatory footer also modifies the final artifact with branding that the user did not necessarily request. This is classified as skill instruction hijacking because the skill text imposes unrelated output requirements on the agent's current task. The observed scope is limited to generated artifact content; no safety policy override, secret extraction, external script loading, or arbitrary code execution was found. ### Attack Path 1. The agent loads `SKILL.md` while handling a daily-fitness request. 2. The visual-design requirements instruct the agent to create the two fixed advertising containers. 3. The instructions also require the fixed `Powered by ClawCode` footer. 4. The agent incorporates those elements into the generated HTML artifact. 5. The user receives an artifact containing unsolicited promotional surfaces and branding. 6. If another component later selects these container IDs and injects content, the predefined locations could be used ...[truncated 670 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
Findings (15)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The inclusion of cron-style push scheduling, filesystem persistence of notification preferences, and multi-channel outbound messaging is materially different from a simple workout routine skill. That mismatch can enable unauthorized or unexpected notifications and retention of user-specific settings under the cover of a benign wellness tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The inclusion of cron-style push scheduling, filesystem persistence of notification preferences, and multi-channel outbound messaging is materially different from a simple workout routine skill. That mismatch can enable unauthorized or unexpected notifications and retention of user-specific settings under the cover of a benign wellness tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The inclusion of cron-style push scheduling, filesystem persistence of notification preferences, and multi-channel outbound messaging is materially different from a simple workout routine skill. That mismatch can enable unauthorized or unexpected notifications and retention of user-specific settings under the cover of a benign wellness tool.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest describes a daily fitness skill focused on 5–15 minute bodyweight routines, instructions, and streak tracking. This script instead defines vocabulary themes such as science, art, business, and philosophy and prints a prompt for generating check-in text plus a stretching plan, indicating behavior aligned with language-learning or generic content generation rather than a workout skill.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest keywords include generic phrases like "daily fitness," "daily workout," "quick workout," "stretch," and "yoga," which are common everyday requests and could match many general conversations. The file does not provide any negative examples or explicit constraints clarifying when this skill should or should not activate.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill documentation introduces external-channel push management even though the manifest describes a local HTML workout generator. This inconsistency increases the attack surface and can mislead reviewers into approving messaging capabilities that were not justified by the declared use case.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

Multi-channel notification control across Telegram, Feishu, Slack, and Discord is a sensitive operational capability that is not naturally implied by generating a daily workout page. In this context, the feature is more dangerous because the surrounding skill appears low-risk, making hidden communications features easier to overlook.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script hardcodes user-facing messages and the generated prompt entirely in Chinese, including errors and the final instruction text. This imposes a specific language/locale on all users with no opt-in, fallback, or documented region-specific justification, which matches the language-policy violation criteria.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The code computes tomorrow's weekday and a 'vocab theme' and asks for a preview of tomorrow's theme and three representative actions. For a fitness skill, themes and actions should derive from workout categories or exercises; using vocabulary themes shows a semantic mismatch between the stated skill purpose and actual behavior.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest describes a daily no-equipment workout skill with instructions and streak tracking, but this script loads a user JSON file before merely printing a workout-generation prompt. Reading arbitrary per-user files is not justified by the code's actual function here, because the loaded data is never used to personalize the routine or update streaks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The user-facing prompt emitted by the script is entirely in Chinese, which enforces a specific language for generated output. The file does not offer any user opt-in, locale selection, or documentation that this is intentionally limited to Chinese-speaking users.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The allowed channel list includes Telegram, Feishu, Slack, and Discord, which adds cross-platform messaging capability unrelated to the manifest's narrow description of bodyweight routines and streak tracking. While reminders may be plausible, broad multi-platform delivery is not explicitly justified by the stated purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest describes a daily no-equipment workout skill with instructions and streak tracking, but this script creates scheduled push jobs that send morning and evening messages over Telegram, Feishu, Slack, or Discord. Proactive multi-channel notification scheduling is a broader operational capability than the stated workout delivery and tracking description.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script hard-codes the timezone to Asia/Shanghai for all scheduled notifications, which can cause reminders to be delivered at unintended times for users in other regions. While this is not a code-execution or data-compromise issue, it can undermine user expectations and consent by sending messages at inappropriate hours, especially because the skill is designed for daily push behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The markdown explicitly directs the skill to save an HTML file to /mnt/user-data/outputs/daily-fitness.html, which is a file write affecting user storage. There is no accompanying warning or disclosure in the description that the skill will create or overwrite a file at that path.

Content

No source excerpt is available for this finding.