The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user-provided fiction text to perform stylistic analysis and rewriting. Malicious instructions hidden within the novel text could potentially influence the agent's behavior during the rewriting process.
Ingestion points: User-provided text snippets or the content of local story files are ingested into the agent's context during the scanning and clearing phases in SKILL.md.
Boundary markers: The skill does not employ specific delimiters or system instructions to distinguish between the story content and agent instructions.
Capability inventory: The skill can execute local shell commands (running Node.js scripts) and perform file system write operations (Edit/Write) to save rewritten content.
Sanitization: Input text is processed without explicit sanitization or filtering to prevent the execution of embedded instructions.
[COMMAND_EXECUTION]: The skill executes local Node.js scripts (scripts/check-ai-patterns.js and scripts/normalize-punctuation.js) to automate punctuation normalization and pattern detection. These scripts are part of the skill package and perform file system read/write operations on targeted story files.

story-deslop