The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands using the gsd-sdk binary via the allowed Bash tool. It interpolates a user-supplied <name> parameter into several subcommands (e.g., create, status, switch, complete).
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an input interpolation security risk surface:
Ingestion points: User input enters the system through the <name> parameter in subcommands defined in SKILL.md.
Boundary markers: Absent; there are no delimiters or instructions guiding the agent to handle special characters or quotes in the input.
Capability inventory: The skill utilizes the Bash tool to execute system commands.
Sanitization: Absent; the user-provided <name> is directly concatenated into shell command strings (e.g., gsd-sdk query workstream.create <name>), which could lead to command injection if an adversarial input contains shell metacharacters.

gsd-workstreams