The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from ROADMAP.md and various phase artifacts to determine its autonomous actions.
Ingestion points: ROADMAP.md, .planning/STATE.md, and phase-specific files (CONTEXT.md, PLANs) in SKILL.md.
Boundary markers: None specified; the agent follows instructions derived from these files.
Capability inventory: Access to Bash, Write, Read, and Agent tools across all execution phases in SKILL.md.
Sanitization: No explicit sanitization or validation of the contents of the roadmap or phase artifacts is described.
[COMMAND_EXECUTION]: The skill executes shell commands via the Bash tool and a vendor-specific CLI tool (gsd-sdk) to initialize milestones and analyze roadmaps.
[DYNAMIC_EXECUTION]: The skill uses Skill() flat invocations to dynamically dispatch and execute phase-specific commands based on discovery at runtime.

gsd-autonomous