The Agent Skills Directory

[COMMAND_EXECUTION]: Automated scans detected that the file examples/trump-perspective/references/research/03-expression-dna.md is infected with HttpRequest-inf [Susp]. This indicates the presence of suspicious network request patterns within the document structure.
[DATA_EXFILTRATION]: The file examples/trump-perspective/references/research/03-expression-dna.md contains a direct link to https://ijels.com/upload_document/issue_files/70IJELS-1042024-TheArt.pdf, which has been identified by scanners as a known botnet URL. Including such links in a skill's research data poses a critical risk of compromising the agent's environment or the user's system.
[COMMAND_EXECUTION]: The file examples/x-mastery-mentor/references/research/05-ai-tech-niche.md has been flagged as FileRepMalware, indicating that its content matches known malware signatures or has a very poor reputation from file scanning engines.
[REMOTE_CODE_EXECUTION]: The skill includes a shell script scripts/download_subtitles.sh that utilizes the yt-dlp tool to download external content based on user-supplied URLs. This execution path allows for arbitrary command execution if the external tool or the downloaded data is manipulated by an attacker.
[EXTERNAL_DOWNLOADS]: The primary operational flow of the skill (Phase 1: Multi-source information collection) involves automated agents performing large-scale data ingestion from untrusted external platforms such as X (Twitter), YouTube, and various blogs. This design creates a massive surface for Indirect Prompt Injection, where malicious instructions embedded in the ingested data could hijack the agent's behavior during the 'distillation' process and compromise any generated skills.

huashu-nuwa