The Agent Skills Directory

[DATA_EXFILTRATION]: The formatBatch function in index.js reads files from user-supplied paths and returns their content to the agent. This allows an attacker to read any file the agent has access to, including sensitive configurations or credentials.
Evidence: index.js lines 343-345 use fs.readFileSync(filePath, 'utf8') and include the results in the function's return object.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing external markdown content without sanitization or protective delimiters.
Ingestion points: markdown parameter in formatMarkdown and file contents in formatBatch (index.js).
Boundary markers: Absent; there are no instructions or delimiters used to separate processed content from agent instructions.
Capability inventory: The skill has local file system read access through the fs module.
Sanitization: Absent; the markdown content is formatted and returned directly without escaping or filtering malicious injection patterns.

markdown-formatter