The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided PDF files to generate summaries and diagrams, which is an inherent ingestion point for untrusted data.
Ingestion points: The book.pdf file (Step 1, Item 3) used as the primary source for content generation.
Boundary markers: The instructions do not define specific delimiters for the PDF text or provide directives to ignore potential instructions embedded within the source material.
Capability inventory: The agent has the ability to write to the local filesystem, execute shell commands via git and mermaid-render, and push updates to a remote repository.
Sanitization: No explicit sanitization or filtering of the content extracted from the PDF is mandated.
[COMMAND_EXECUTION]: The skill utilizes specific shell commands to manage the repository and validate generated diagrams.
Evidence: Commands include git clone, git push, and mermaid-render /tmp/d.mmd (Steps 0, 2, and 4).
Context: These tools are authorized in the skill's metadata and are employed for expected administrative tasks within the scope of the project.
[EXTERNAL_DOWNLOADS]: The skill performs a repository clone operation from an external service.
Evidence: git clone git@github.com:mmontes11/knowledge-base.git (Locate the knowledge base section).
Context: The repository is the author's own official knowledge base, representing a legitimate vendor resource.

knowledge-base-add-book