The Agent Skills Directory

[SAFE]: The skill's primary function is to detect the current project's technology stack (e.g., reading package.json, go.mod) and initialize a configuration directory (openspec/) or save the context to a memory system (engram).
[COMMAND_EXECUTION]: The skill performs directory and file creation (e.g., creating openspec/ and .atl/ folders). These are standard project initialization tasks and do not involve arbitrary or dangerous command execution.
[INDIRECT_PROMPT_INJECTION]: The skill scans for project-level instructions such as CLAUDE.md, .cursorrules, and agents.md. While this ingests external data into the agent's context, the skill uses these specifically to build a 'skill registry' and does not provide an automated execution path for instructions found within those files without further agent processing. This is a standard architectural feature for agentic workflows.
[DATA_EXPOSURE]: The skill reads project configuration files and environment metadata to determine the stack. It persists this context locally or to a designated memory backend (mem_save). There is no evidence of data being exfiltrated to untrusted external domains; persistence is confined to the user's controlled environment.

sdd-init