The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-supplied Goodreads CSV exports, creating a surface for indirect prompt injection where malicious instructions could be embedded in book reviews or metadata.
Ingestion points: SKILL.md (Section 2) and references/csv-reconciliation.md describe the process of parsing the user-provided CSV file.
Boundary markers: The skill requires a "preview of the exact write set" and "explicit approval before any writes." It also uses specific HTML comment markers to delimit imported content from human-written prose.
Capability inventory: The skill possesses the capability to write and update Markdown files and frontmatter within the Obsidian vault (SKILL.md Section 4).
Sanitization: references/csv-reconciliation.md mandates the use of a safe CSV parser, forbids the evaluation of spreadsheet expressions to prevent CSV injection attacks, and requires "safe HTML-to-Markdown conversion" that excludes scripts or remote content fetching.

import-goodreads