The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes two Node.js scripts, analyze-migration.js and scaffold-feature-move.js, which are designed to be executed locally. These scripts have the capability to read directories, create new folders, move files, and append content to files across the user's project structure.
[COMMAND_EXECUTION]: Multiple TypeScript template files located in assets/feature-template/ are encoded using UTF-16 Little Endian (with a Byte Order Mark). This encoding method obscures the file contents from simple text-search tools and security scanners that primarily check for UTF-8 or ASCII strings, effectively acting as a form of code obfuscation.
[DATA_EXFILTRATION]: The scaffold-feature-move.js script lacks path sanitization for the featureName and layer arguments. This allows for potential path traversal, where a malicious or accidental input could move sensitive system files (e.g., credentials or configuration) into the project's source folders where they might be exposed or committed to version control.
[PROMPT_INJECTION]: The analyze-migration.js script identifies a vulnerability surface for indirect prompt injection by processing untrusted file system data.
Ingestion points: fs.readdirSync and fs.readFileSync in scripts/analyze-migration.js ingest file paths and names from the ./src directory.
Boundary markers: None. The script does not use delimiters or instructions to prevent the agent from interpreting filenames as instructions during the audit process.
Capability inventory: The skill allows for file movement and file content modification via the accompanying scaffold-feature-move.js script.
Sanitization: None. Filenames and paths are used directly in output reports and passed as arguments to the scaffolding script without validation or escaping.

fdd-architecture