daymark-development

name	daymark-development
description	Develop, refactor, diagnose, or review the Daymark Flask, SQLite, worker, MCP, monitor, report, project, meeting, agent-runner, and UI code. Use for changes in the Daymark repository that must preserve isolated production state and follow its package boundaries, migration rules, and verification workflow.

Daymark Development

Use this skill for repository work on Daymark. Read AGENTS.md first, then consult docs/architecture.md for package ownership. Load references/verification.md when the task changes code, packaging, migrations, or deployment. Keep the generated Codex interface metadata in agents/openai.yaml aligned with this workflow.

Work safely

1. Inspect git status before editing and preserve unrelated changes.
2. Work only in the development checkout. Never mutate or restart the production checkout unless the user explicitly authorizes deployment.
3. Keep runtime data, copied databases, credentials, logs, PID files, and generated agent workspaces outside Git.
4. Use a temporary DAYMARK_DATA_DIR for runtime and migration checks.

Place changes

• Put HTTP and UI coordination in daymark.web.
• Put durable operations and explicit forward migrations in daymark.storage.
• Put harness-neutral model execution in daymark.ai; select harness and model through configuration.
• Put MCP tools in the matching module under daymark.mcp.
• Put domain behavior in projects, meetings, monitors, reports, or journal, keeping Flask and SQLite mechanics at their boundaries.
• Add package assets under daymark/templates or daymark/static.

Prefer focused modules and public domain operations over route-local logic, cross-package private imports, or compatibility shims.

Verify and hand off

Run focused tests while iterating, then complete the checklist in references/verification.md. Report which checks passed and any unverified external behavior. When publication is requested, commit only intentional changes, push the feature branch, and open a draft pull request with the behavioral impact and verification results.