kerno-plan-implement-baseline

Kerno MCP — plan and implement scenario baselines

When to use this skill

Use kerno_plan_baseline and kerno_implement_baseline when the goal is to create or update .scenario.ts files and confirm they pass end-to-end.

Typical situations:

• Scenarios are missing for routes (check existingTests via kerno_list_endpoints with the same scope).
• kerno_validate failed and scenarios need Kerno-driven regeneration — do not hand-edit .kerno/scenarios/.
• The user wants a batch pass (scope: all) or a single route (scope: endpoint:… or sync method + path).

Do not use this skill as the primary guide for:

• kerno_validate alone — use skills/kerno-validate/SKILL.md. Run validate before deciding scenarios need regeneration when the API changed.

Prerequisites

1. kerno_healthcheck and kerno_get_applications (note workspace_id if you will pass it).
2. Optional kerno_list_endpoints with the same scope you will use for plan/implement — requires scope (e.g. all).
3. kerno_implement_baseline only: stack up with ready_for_validation: true (via kerno_compose_up or kerno_start_environment + kerno_job). Planning (kerno_plan_baseline) does not require a running environment.

Use the same workspace_path and app conventions as compose and start-environment tools.

Stage 1 — Plan (kerno_plan_baseline)

Async (many endpoints): call with scope (not changed). Optional plan_review_instructions. Returns job_id, log_path, kind: plan_baseline. Complete with kerno_job.

Sync (one endpoint in the MCP loop): omit scope; pass method and path together. Optional plan_review_instructions on each round. Response has plan_sync — no kerno_job for that call.

Present the plan to the user (scenario names, coverage, open decisions). Refine with another sync call or async job + plan_review_instructions before implement.

Stage 2 — Implement (kerno_implement_baseline)

Only after planning for every endpoint in scope and ready_for_validation is true:

1. Call kerno_implement_baseline with scope (and optional app, workspace_id).
2. Complete with kerno_job (kind: implement_baseline).

Long-running LLM work — use sparse wait=false polling or log_path. See skills/kerno-background-job/SKILL.md.

Scopes

Scope	Plan (async)	Implement
all	yes	yes
file:<path>	yes	yes
endpoint:METHOD /path	yes	yes
changed	no — use kerno_validate	no

Parameter details and terminal JSON: ${CLAUDE_PLUGIN_ROOT}/references/plan-implement-baseline.md.

Deduping

plan_baseline and implement_baseline share one slot per workspace + scope + app selection. Do not launch a second job in that family while one is running.

See also

• ${CLAUDE_PLUGIN_ROOT}/references/tool-ordering.md — full tool order.
• ${CLAUDE_PLUGIN_ROOT}/skills/kerno-validate/SKILL.md — validate before regenerating when the API changed.
• ${CLAUDE_PLUGIN_ROOT}/references/compose-plan.md — environment setup before implement.