| name | hecate-architect |
| description | Use when planning a substantial change before coding — new wire fields, new persisted things, new endpoints, new persistent UI surfaces, cross-package refactors. Produces a structured plan, not code. |
Hecate architect skill
Plan-shaped responses for substantial changes. The skill produces a plan, not code. Use it when "I am about to make a substantial change; I need to think before I type."
When to use
Any change that triggers a planning event in ../../core/workflow.md:
- Cross-package wire-field changes (the seven-step chain — see
../providers/SKILL.md).
- New Hecate-owned persisted things — normally mirror memory + SQLite + Postgres
tiers. Portable Projects records belong in Cairnline instead; never add a
second Hecate persistence tier for them.
- New HTTP endpoints (provider-compatible
/v1/... or Hecate-native /hecate/v1/...).
- New approval policies or sandbox capabilities.
- New persistent UI surfaces (inspector, side rail, dashboard block, summary panel).
- Substantive refactors that cross ring boundaries or touch the api↔providers seam.
When in doubt, default to using this skill. The cost of a brief plan is far less than the cost of an unplanned change that hits the wrong seam.
Bias
Plan first. Do not write code in this skill's response. Unless the task is trivial (typo fix, single-line dependency bump, comment touch-up), produce a plan first.
Output
Use the canonical plan structure and anti-patterns in ../../tasks/planning.md. That doc owns the format; this skill owns when to invoke it.
Hand-off
The plan is the brief for a hecate-backend, hecate-ui, or hecate-providers skill turn. It should give the implementer everything they need to proceed without further context — including the verification ladder they'll run and the docs they'll update.
Verification expectations
None for the plan itself. The plan must enumerate the verification ladder for the implementer.
