The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard system tools such as mkdir, grep, and rg, as well as specific CLI tools like nmem (for memory) and obsidian (for vault access). These operations are restricted to the local repository and the user's defined Obsidian vault, which is typical for a workspace-integrated writing assistant.- [DATA_EXFILTRATION]: While the skill accesses user content (representative works and personal memories) and performs web searches, it does so to fulfill its primary purpose of style-consistent writing. It includes explicit instructions to de-identify sensitive data and handle privacy risks, and it does not demonstrate any hidden or unauthorized data transmission patterns.- [INDIRECT_PROMPT_INJECTION]: As a writing tool that processes external data (web search results and user samples), the skill has an inherent surface for indirect prompt injection. However, it implements a comprehensive 'Gate Check' system, 'Claim Ledger' for fact verification, and 'Linting' processes to audit output before finalization, effectively mitigating this risk.- [OBFUSCATION]: A thorough review of all 40 files (including the provided SKILL.md and reference files) revealed no use of Base64 encoding for hidden commands, zero-width characters, or other obfuscation techniques.- [SAFE]: The skill architecture follows best practices for AI agents, emphasizing modularity, user confirmation for style updates, and evidence-based content generation. All external dependencies (CLI tools) are identified as part of the intended local development environment.

writing-clone-profile