The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface identified in the ingestion of project documentation.
Ingestion points: Reads files from production/milestones/, production/sprints/, design/gdd/, and production/risk-register/ (SKILL.md).
Boundary markers: Missing; the agent interpolates external document content directly into its reasoning process.
Capability inventory: Includes Write, Edit, and the Task tool for spawning sub-agents (SKILL.md).
Sanitization: None; external text is not validated or escaped before processing.
[COMMAND_EXECUTION]: Execution of shell commands via dynamic context injection at load time.
Evidence: !ls production/sprints/ 2>/dev/null in the frontmatter context field.
Analysis: Used for benign discovery of existing sprint files to populate the agent's context. Does not include user-controllable arguments.
[COMMAND_EXECUTION]: Utilization of the Task tool for agent-to-agent delegation.
Evidence: Spawns a producer agent using the PR-SPRINT gate (Phase 4).
Analysis: Standard multi-agent orchestration pattern for reviewing project feasibility.

sprint-plan