The Agent Skills Directory

[DATA_EXFILTRATION]: The skill references a sensitive file path, C:/Users/<user>/.claude/CLAUDE.md, for modification. This path points to a configuration file within the user's profile directory on Windows. Instructing an agent to access or alter hidden configuration files in user directories is a sensitive operation that exposes local system structure and environment settings.
[COMMAND_EXECUTION]: The skill orchestrates multiple file system operations, including creating directories and writing files. It uses a user-supplied name variable to define the target directory path (docs/members/the-<name>/). Without validation, this creates a path traversal vulnerability where an attacker could provide a malicious string (e.g., ../../) to cause the agent to write files to unauthorized locations outside the project directory.
[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection by collecting user-provided descriptions and logic to populate skill templates.
Ingestion points: Data enters through user-provided names, specialties, and philosophical descriptions during the member creation process in SKILL.md.
Boundary markers: Absent; there are no instructions to use delimiters or warnings to ignore instructions embedded within the user-provided data.
Capability inventory: The skill possesses the capability to write and modify files within the repository and the user's local configuration.
Sanitization: Absent; the skill does not require validation or escaping of the user-provided content before it is interpolated into the new SKILL.md and README.md files.

institutional-knowledge