The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it ingests untrusted data from external planning artifacts and epic files to generate implementation tracking.
Ingestion points: The scripts/sprint_plan.py script reads Markdown epic files, and the readiness-gate.md instructions direct the agent to scan {planning_artifacts} and {project_knowledge}.
Boundary markers: No explicit delimiting or instructions to ignore embedded commands are used when reading these artifacts.
Capability inventory: The skill has the capability to write to the file system (sprint-status.yaml) and influence the agent's recommendations for subsequent development steps.
Sanitization: The script performs basic sanitization of story titles using the _slug function to ensure keys are safe for YAML structure.
[SAFE]: The Python script scripts/sprint_plan.py uses the ruamel.yaml library with the RoundTripLoader (typ='rt'). This is a safe loading mechanism that does not execute arbitrary code, making the static analysis alerts for unsafe YAML loading false positives in this context.

bmad-sprint-planning