The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute gsd_run for parsing arguments and managing the batch workflow. User-controlled data from $ARGUMENTS is interpolated into these shell commands. The instructions specify quoting the input to prevent shell metacharacter expansion.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a task coordinator that ingests and processes multi-item task lists, creating a surface for indirect prompt injection when this data is passed to downstream agent leaves.
Ingestion points: Untrusted data enters the agent context through the $ARGUMENTS variable and external files referenced via the --file flag in SKILL.md.
Boundary markers: The skill explicitly requires wrapping task descriptions in DATA_START/DATA_END markers and using <security_context> blocks to define the content as untrusted data for sub-agents.
Capability inventory: The skill has access to Bash, Agent, Read, Write, Edit, Glob, and Grep tools.
Sanitization: The skill employs shell quoting for interpolated arguments and routes validation through a dedicated CLI tool (quick-batch parse-args) before further processing.

gsd-quick-batch