The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied arguments and project files, which represent an attack surface for indirect instructions.
Ingestion points: User input via the $ARGUMENTS variable and project phase metadata from ROADMAP.md mentioned in SKILL.md.
Boundary markers: The skill uses XML-style tags (<arguments>, <objective>) to delimit data from instructions.
Capability inventory: The skill possesses extensive capabilities including Bash, Write, Read, and the ability to spawn sub-agents via the Agent tool.
Sanitization: Instructions explicitly direct the agent to treat the input block as data rather than instructions, providing a prompt-level guardrail.
[COMMAND_EXECUTION]: The skill is designed to execute shell-based operations and interact with local CLI tools.
Evidence: The skill uses the Bash tool and references vendor-specific utilities such as gsd-tools, gsd-plan-phase, and gsd-plan-review-convergence to manage milestone operations and phase analysis.

gsd-autonomous