The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands (ln, cp, readlink) to manage configuration files and symlinks within the user's config directory.
[DATA_EXPOSURE]: The skill accesses and displays the contents of vault configuration files which are likely to contain sensitive information like tokens or paths.
The Show command instructs the agent to print file contents verbatim.
A safety mitigation is included: "redact any lines containing API_KEY or SECRET — show *** instead of the value".
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied vault names directly into shell command templates, creating a potential injection surface.
Ingestion points: User-provided vault names in commands like /wiki-switch <name> or /wiki-switch new <name> (SKILL.md).
Boundary markers: None present for the user-supplied name variable.
Capability inventory: Execution of ln, cp, and file read/write operations (SKILL.md).
Sanitization: Includes a specific redaction rule for API_KEY and SECRET in the Show command, but lacks explicit sanitization or validation for shell command arguments like <name>.

wiki-switch