The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external file which could potentially contain malicious instructions or unexpected paths if the source is compromised.
Ingestion points: Reads the path property from workspace.json (located in ~/.local/share/claude-plugins/image-production/ or similar) in SKILL.md.
Boundary markers: Absent. The script does not include specific delimiters or instructions to the agent to ignore content within the path variable.
Capability inventory: The skill has the capability to execute shell commands (ls, konsole, gnome-terminal, xdg-open) via the Bash tool as seen in SKILL.md.
Sanitization: The script uses test -d "$WS" to verify the path is a valid directory before execution, which provides a basic layer of validation.
[COMMAND_EXECUTION]: The skill intentionally executes shell commands to facilitate user workflows by opening terminal emulators or file managers at a specific path.
Evidence: The script in SKILL.md conditionally executes konsole, gnome-terminal, or xdg-open based on their availability on the system.

open-workspace