The Agent Skills Directory

[SAFE]: The skill uses local Python scripts and Git for project state management and chapter backups, which are appropriate and expected behaviors for a creative writing assistant. No untrusted external dependencies or remote code downloads were found.- [COMMAND_EXECUTION]: Shell commands are used to execute local project scripts and standard Git operations. These commands rely on environment variables that reference internal plugin and project paths.- [PROMPT_INJECTION]: No malicious instructions intended to bypass safety guardrails or extract system prompts were detected. The workflow uses modular agents with specific, safe prompts for creative writing tasks.- [DATA_EXFILTRATION]: The documented Git backup feature allows users to preserve their work. No evidence of unauthorized data transmission to external or untrusted domains was found.- [PROMPT_INJECTION]: The skill processes external project files which presents a theoretical surface for indirect prompt injection.
Ingestion points: Project configuration and outlines are read from files like .webnovel/state.json and MASTER_SETTING.json into the agent's context.
Boundary markers: The skill uses structured key-value assignment in prompts when calling subagents via the Agent tool.
Capability inventory: The agent has access to Read, Write, Edit, Grep, Bash, and Agent tools across the workflow.
Sanitization: While no explicit sanitization is detailed, the use of specialized subagents with narrow task scopes (e.g., reviewer, data-agent) serves to isolate the processing of untrusted content.

webnovel-write