The Agent Skills Directory

[METADATA_POISONING]: The skill's YAML frontmatter identifies the author as 'https://github.com/dmitry', which is inconsistent with the provided author context of 'jeffallan'. This discrepancy in metadata can be misleading regarding the skill's origin.
[INDIRECT_PROMPT_INJECTION]: The skill processes content derived from external research sources, creating a potential surface for indirect prompt injection where malicious instructions could be embedded in the ingested data.
Ingestion points: Research artifacts, synthesis notes, and source data identified during upstream research and writing phases (described in SKILL.md and references/harvest-process.md).
Boundary markers: The skill explicitly requires human approval via the AskUserQuestion tool before any data is written to the file system, acting as a critical control point (SKILL.md, Constraints).
Capability inventory: The skill has the capability to write markdown files to a user-defined vault path on the file system (SKILL.md, Core Workflow).
Sanitization: Ingested content is formatted into structured Markdown and YAML frontmatter templates, providing structural consistency and preventing raw data injection into sensitive file formats, although semantic filtering is not performed (references/vault-format.md).

knowledge-harvester