The Agent Skills Directory

[SAFE]: The skill consists entirely of Markdown instructions and reference files. No executable scripts, binaries, or automated package installations (Python or Node.js) are included.
[SAFE]: The skill provides technical references to academic and open-source Natural Language Processing (NLP) research projects from reputable sources (e.g., NILC/USP). These links are informational and do not involve remote code execution or suspicious downloads.
[PROMPT_INJECTION]: No malicious prompt injection or behavior override patterns were detected. The instructions reinforce factual consistency and safety by directing the agent to maintain the source text's arguments, data, and modal context without invention.
[DATA_EXFILTRATION]: No hardcoded credentials, sensitive file path access (such as .ssh or .env), or unauthorized network operations were identified. External URLs are restricted to the author's official repository and legitimate research documentation.
[COMMAND_EXECUTION]: The skill does not utilize dynamic context injection (shell commands using the exclamation-backtick syntax) or any other form of runtime command execution.
[SAFE]: While the skill ingests user-supplied text for processing, it lacks tools or code execution capabilities that could be exploited via indirect prompt injection. The instructions include explicit boundary markers and factual integrity requirements.

humanizar