The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from documents and URLs, which represents a surface for indirect prompt injection. Maliciously embedded instructions could attempt to influence the memory extraction. This risk is addressed by the use of predefined templates, authority rankings for sources, and a structured integration checklist.
Ingestion points: SKILL.md and references/extraction-rules.md facilitate reading various document formats and web content.
Boundary markers: The skill uses templates for organizing information but does not define strict data-instruction delimiters for the source text.
Capability inventory: Reading local files and URLs; writing markdown files; executing the provided init_memory_pack.py script.
Sanitization: No specific content sanitization or instruction filtering is performed on the ingested text.
[COMMAND_EXECUTION]: The workflow utilizes scripts/init_memory_pack.py to set up the memory pack directory structure. This script is distributed with the skill and uses standard libraries to create and write markdown templates without external dependencies.

document-memory-builder