The Agent Skills Directory

[DATA_EXFILTRATION]: The skill references specific absolute local file paths, which exposes the host's directory structure and user profile information.
Evidence: references/source-map.md contains multiple hardcoded absolute paths under the /Users/braden/Code/epicenter-md/ directory, revealing the specific organization and location of the author's local files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external essay drafts and notes, but does not provide instructions for handling potentially malicious content embedded within those files.
Ingestion points: The agent is directed to read specific files (drafts, course notes, and prompts) listed in references/source-map.md when critiquing writing.
Boundary markers: Absent. The instructions do not specify the use of delimiters or provide warnings to the agent to ignore any embedded instructions found within the processed text files.
Capability inventory: The agent has the capability to draft revisions, provide critiques, and interpret writing principles, which could be influenced if a document contains instructions disguised as prose.
Sanitization: Absent. There are no instructions to sanitize, validate, or filter the content of the linked files before the agent processes them.

english-120-writing