The Agent Skills Directory

[SAFE]: The skill functions as a high-level control plane that delegates concrete actions to specialized sub-skills and tools. It incorporates robust security practices such as configuration persistence in a dedicated directory and mandatory user approval for all phase transitions.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to perform repository maintenance via the references/migration.md guide. This involves the use of standard commands like git mv, rmdir, and grep. These operations are scoped strictly to the migration of documentation files and are gated by a requirement for the agent to present a migration plan for user approval before execution.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests data from external sources such as Jira tickets, Linear issues, and Confluence pages to determine workflow state. This risk is mitigated by the architectural enforcement of 'Approval Gates' where the user must verify the agent's interpretation and proposed actions at every step.
Ingestion points: Project documentation files in the documentation/ folder and external tracker work-items (Jira, Linear, Azure DevOps).
Boundary markers: None explicitly defined in the orchestrator logic.
Capability inventory: Repository file manipulation (git mv, rmdir), tracker status transitions, and orchestration of code-executing sub-skills.
Sanitization: Not explicitly implemented; relies on human verification at approval gates.

ol-sdd-workflow