The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a suite of local scripts (author_memory_commit.py, tracking_commit.py, check-ai-patterns.js, normalize-punctuation.js, and check-degeneration.js) to manage story state, perform word counts, and scan prose for stylistic issues. These scripts are provided within the skill bundle and use standard libraries for local file operations without external network dependencies.
Evidence: Script execution instructions in SKILL.md and the corresponding source code in the scripts/ directory.
[INDIRECT_PROMPT_INJECTION]: As a text analysis tool, the skill ingests user-provided novel content which is then processed by LLM sub-agents. This presents an attack surface for indirect prompt injection where instructions embedded in the fiction text could attempt to bias the review. The skill mitigates this by using modular sub-agents with focused instructions and performing structural validation on story metadata.
Evidence: Instructions in SKILL.md Phase 1 and 2 detail the ingestion of novel files and interpolation of content into sub-agent prompts.
Mitigation: Sub-agents are coordinated in a serial, controlled manner with specific evaluation rubrics to maintain focus on the review task.

story-review