The Agent Skills Directory

[SAFE]: The skill defines a sophisticated and transparent workflow for document processing and knowledge extraction. The logic is clearly documented across various methodology and extractor files.
[SAFE]: No malicious command execution, unauthorized data exfiltration patterns, or obfuscation techniques were found in the instructions or templates.
[SAFE]: All external repository references and links point to the vendor's own infrastructure or to explicitly mentioned collaborative projects, consistent with the ecosystem described in the documentation.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing untrusted user-provided text (books), which constitutes an indirect prompt injection surface.
Ingestion points: The skill (SKILL.md) reads book content from user-provided file paths in formats such as PDF, EPUB, and TXT.
Boundary markers: The workflow incorporates a 'Triple Verification' phase (methodology/03-stage1.5-triple-verify.md) which includes cross-domain validation and predictive power tests. These act as logical filters for extracted content, though no technical prompt delimiters (e.g., XML tags or special tokens) are explicitly used to wrap the raw source text during processing.
Capability inventory: The skill utilizes sub-agent spawning for parallel extraction and performs file system writes to save the distilled skill modules (SKILL.md, extractors/framework-extractor.md).
Sanitization: There is no explicit sanitization or filtering logic designed to detect or neutralize malicious instructions embedded within the source book text.

book2skill