The Agent Skills Directory

[SAFE]: Analysis of the skill instructions and structure indicates safe behavior consistent with its stated purpose of creating implementation task lists.
[DATA_EXFILTRATION]: The skill interacts with the local file system to read design documents and write task lists within the openspec/ directory. These operations are scoped to the project environment and do not involve sensitive system files or credentials.
[PROMPT_INJECTION]: The skill processes external data including proposals, specs, and design documents. While this constitutes a potential surface for indirect prompt injection, the risk is mitigated by the skill's limited output format (markdown checklists) and lack of high-privilege execution or network capabilities.
Ingestion points: proposal.md, specs/, design.md in openspec/changes/{change-name}/
Boundary markers: Absent
Capability inventory: File-write to tasks.md
Sanitization: Absent

sdd-tasks