The Agent Skills Directory

[DATA_EXFILTRATION]: The conversion script resolves and reads local files based on image paths provided in the Markdown input (![alt](path)). This mechanism could be used to read arbitrary files from the filesystem if a malicious Markdown file is provided.
Evidence: The script uses resolve(inputDir, href) followed by readFileSync(imgPath) in scripts/md-to-docx.mjs to embed images.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Markdown files which could contain hidden instructions intended to manipulate the agent's behavior during the conversion process.
Ingestion points: The inputPath file is read and parsed by the marked library in scripts/md-to-docx.mjs.
Boundary markers: Absent. The skill does not provide instructions to the agent to ignore embedded instructions in the source file.
Capability inventory: The skill can execute shell commands (npm install, node), read files from the workspace, and write output files.
Sanitization: No sanitization or filtering is performed on the Markdown content to prevent instruction injection.
[COMMAND_EXECUTION]: The skill instructions require the agent to execute shell commands to install dependencies and run the Node.js converter. While standard for this tool type, it involves executing code defined in the local package.json.
Evidence: SKILL.md contains instructions for npm install and executing the md-to-docx.mjs script via Node.js.

md-to-docx