other
Warning
- Location
- skill.md:523
- Finding
- Automatic User-Linked Telemetry Without Documented Consent or Destination## Vulnerability Details **File Location**: `skill.md`, lines 523-571 **Vulnerability Type**: Undisclosed telemetry and collection of user-linked usage data **Risk Level**: Medium ### Vulnerable Code ```javascript // Skill startup trackEvent("skill_start", { skill_id: "content-adapter", user_id: userId, source: "workbuddy", input_type: "text" }); // Skill completion trackEvent("skill_complete", { skill_id: "content-adapter", user_id: userId, output_format: "html", output_size: 45000, total_duration_ms: 3200, word_count: 1200, platform: "xiaohongshu" }); ``` The surrounding instructions state that events are automatically reported at Skill startup, after generation steps, upon completion, on errors, and when users share generated content. They also state that analytics failures are silent and do not interrupt normal execution. ### Technical Analysis The Skill's legitimate purpose is local transformation of user-provided content into platform-specific writing styles. That task does not require a persistent user identifier or remote reporting of user behavior. The telemetry specification collects or proposes collecting: - A user identifier - Invocation source and input type - Step names and execution timings - Output format, size, and word count - Target publishing platform - Sharing platform and content type - Error information No analytics endpoint, data controller, consent mechanism, retention period, identifier-pseudonymization policy, or opt-out procedure is provided. The referenced `analytics-sdk` implementation is also absent from the audited package, preventing verification of its transmission destination and safeguards. The package does not contain executable telemetry code, so successful data exfiltration cannot be confirmed from the audited files alone. The confirmed issue is that the Skill specification directs automatic, user-linked reporting beyond ...[truncated 1357 chars]
- Remediation
- ## Remediation Suggestions 1. Remove telemetry because it is not necessary for the Skill's content-transformation function. 2. If analytics is retained, require explicit opt-in consent before collecting or transmitting any event. 3. Do not collect a direct `user_id`; use a short-lived, randomized, non-linkable identifier where metrics genuinely require one. 4. Publish the exact destination endpoint, data controller, event schema, retention period, legal basis, and deletion procedure. 5. Provide a clearly documented opt-out that disables all analytics calls. 6. Display reporting failures and configuration status rather than intentionally suppressing all telemetry errors. 7. Include the referenced SDK source or a pinned, verifiable dependency so its network behavior can be audited. 8. Restrict events to aggregate operational metrics and exclude publishing targets, sharing activity, input characteristics, and error content unless strictly required.