The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests structured and prose results from specialized subagents (Executor, QA) to maintain flow state and build Fix Plans for addressing issues.
Ingestion points: Worker results are stored in worker_results, executor_results, and qa_results as described in the "Flow State" section of SKILL.md.
Boundary markers: The skill relies on markdown headers (e.g., ## STATUS, ## WORK COMPLETED) to delimit worker data but does not explicitly warn subagents to ignore instructions embedded within those sections.
Capability inventory: The orchestrator has access to Bash, Read, Glob, and Grep tools, and it has the capability to launch and parameterize new subagents.
Sanitization: There is no mention of sanitizing or escaping the content provided by workers before it is interpolated into the briefs for subsequent workers.
[DYNAMIC_EXECUTION]: The skill dynamically assembles instruction "briefs" for subagents based on the current flow state and prior results.
Evidence: The "Worker Launch Pattern" and Steps 1-4 in SKILL.md detail how the orchestrator constructs prompts (briefs) containing prior results and state data to be executed by new subagents.
[COMMAND_EXECUTION]: The skill uses shell commands to check the status of the project environment.
Evidence: The Workflow Step 0 in SKILL.md specifies running kyro status via the Bash tool to identify the active sprint and stage.

kyro-sprint-flow