The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides stylistic instructions to improve writing quality. There are no attempts to bypass safety filters, override system prompts, or utilize jailbreak patterns.
[DATA_EXFILTRATION]: No hardcoded credentials, sensitive file path access (e.g., .ssh, .env), or network exfiltration logic was detected within the skill instructions or metadata.
[REMOTE_CODE_EXECUTION]: The skill does not perform any runtime downloads or execute external scripts. The installation instructions in the README are standard, user-initiated Git operations and do not represent automated remote code execution.
[OBFUSCATION]: Analysis of the markdown and YAML content revealed no hidden characters, Base64-encoded payloads, or homoglyph-based URL spoofing.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and rewrite user-provided text. While this is an ingestion surface, the instructions focus on linguistic pattern matching and lack unsafe interpolation or high-privilege tool chains that could be exploited via malicious input data.
[COMMAND_EXECUTION]: The skill uses allowed tools (Read, Write, Edit) for its intended purpose of text manipulation. No dangerous system command execution or shell injection vectors were identified.

humanizer