The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs users to navigate to x.com and paste code into the browser's DevTools console. This execution method bypasses security boundaries like Same-Origin Policy (SOP) and Content Security Policy (CSP), allowing the scripts to act with the full permissions of the logged-in user.
[NO_CODE]: The skill relies on 18 separate JavaScript files (e.g., src/engagementAnalytics.js, src/auditFollowers.js) that are not included in the skill package. Directing users to execute missing or external code is a significant security risk as the logic cannot be audited for malicious behavior.
[METADATA_POISONING]: The skill metadata lists the author as 'nichxbt', which contradicts the provided author identity 'nirholas'. This discrepancy indicates a potential impersonation attempt to establish false credibility.
[INDIRECT_PROMPT_INJECTION]: The skill's stated purpose is to ingest and analyze data from X/Twitter, such as timelines, hashtags, and follower profiles. This introduces a vulnerability surface where malicious instructions could be embedded in the social media content being analyzed.
Ingestion points: Browser scripts scrape content from x.com timelines, search results, and follower lists.
Boundary markers: There are no markers or instructions provided to distinguish between data and control instructions during the analysis process.
Capability inventory: Scripts have full DOM access and can perform network operations via the browser's fetch or XMLHttpRequest APIs.
Sanitization: No sanitization or filtering logic is described for the content being scraped, potentially allowing malicious data to influence the agent's behavior or the user's environment.

analytics-insights