The Agent Skills Directory

[SAFE]: The skill implements a multi-perspective review process for creative writing. It provides structured instructions for analyzing story structure, character development, and prose quality based on platform-specific rubrics.\n- [SAFE]: The skill bundles several local scripts to automate quality assurance: \n
scripts/normalize-punctuation.js, scripts/check-ai-patterns.js, and scripts/check-degeneration.js use Node.js to scan and normalize story text. \n
scripts/review_tickets.py and scripts/tracking_commit.py use Python to manage a structured project state, including character snapshots and plot timelines, stored in .moshu-review/ and 追踪/.\n- [SAFE]: A security mechanism using "Review Tokens" (审稿令牌) is employed to verify that sub-agents have processed the actual input content, preventing hallucinations or report fabrication.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external story text. \n
Ingestion points: The agent reads novel content from the 正文/ directory and outlines from 大纲/. \n
Boundary markers: Instructions explicitly state that the goal is identifying problems rather than verifying correctness, and use a structured "Findings Schema" to parse results. \n
Capability inventory: The skill can execute local scripts to modify files within the project directory and spawn sub-agents for specialized tasks. \n
Sanitization: The skill focuses on pattern detection and structured reporting rather than interpreting the data as executable instructions.

moshu-review