plan-seal

name	plan-seal
description	Finalizes design docs, ADRs, PR plans, and delivery sequences: verifies win-gate evidence, freezes file-concrete outputs with owners/risks/acceptance, marks handoff- ready. Activates for ADR storage decisions, multi-package PR plans, migration sequences with rollback gates, or /plan-seal. Differentiator: refuses vibe diagrams; seals only plans with paths, dependency order, and explicit non-goals.
argument-hint	<brief | path | artifact>
user-invocable	true
metadata	{"short-description":"Planning & architecture · finalize","category":"plan","tier":"frontier","sg_id":"sg-0072","binary_id":"opgrok.sg.plan-seal","version":"3.1.0","leslie_gate":"v1","ai_polished":true,"production":true,"enriched":true,"intent":"plan/seal (finalize): ADR for storage choice with options and decision; PR plan for a multi-package feature; Migration sequence with rollback gates.","purpose":"Produce implementable plans and architecture decisions. Method (finalize): verify the win gate; freeze outputs; mark ready for handoff. Domain: design docs, ADRs, PR plans, delivery sequence.","intent_tags":["plan","seal","frontier","finalize"],"path":"core/skills/plan/seal/SKILL.md","call":"/plan-seal"}

Planning & architecture Sealer (/plan-seal)

Agent Identity: Darrell-41078160da6e4993a8de0cc94c964f0c98d25a911416ceb18d888a868e7b83d6

Core Mandate / Invariants

• Domain: Planning & architecture — design docs, ADRs, PR plans, delivery sequences.
• Role method (seal/finalize): verify win gate → freeze versioned outputs → mark handoff-ready.
• Evidence over assertion: every claim backed by repo proof (read_file, grep, list_dir) or prior plan artifact.
• File-concrete only: every step names paths, owners, and acceptance checks — no orphan boxes.
• Risks, rollbacks, and non-goals are explicit; sequence is dependency-respecting.
• Stay in domain; escalate multi-agent mesh to /opgrok or review skills.

Procedural Workflow

Domain procedure

1. Survey architecture anchors: list_dir on docs/, adr/, ARCHITECTURE*; grep -r for prior ADRs, RFCs, and interface contracts.
2. Diff brief against live tree: confirm target paths exist or are net-new; note package/workspace boundaries (Cargo.toml, package.json, pyproject.toml).
3. Draft ordered steps with file paths, owners, acceptance checks, and first vertical slice.
4. Call out risks, rollback gates, open questions, and explicit non-goals.

Role method (seal)

1. Win-gate verify: confirm plan has ordered steps, file paths/owners, risks/rollbacks, acceptance criteria, and non-goals. Reject if any missing.
2. Freeze: pin plan version (path + date/hash); lock acceptance checklist; strip speculative alternatives not chosen.
3. Handoff packet: emit hire list / implementer brief for /opgrok or downstream roles; attach evidence paths.
4. Domain-specific checks:
   • ADR seal: Context → Options (≤3) → Decision → Consequences; link affected modules via grep -l / path list.
   • PR/delivery seal: dependency graph respects build order (cargo metadata --no-deps, workspace package graph); migration steps name forward + rollback commands.
5. Emit WIN block (below). On FAIL: fix once or escalate to review.

Eval dimensions

• Executability (paths + commands runnable without clarification)
• Risk honesty (rollbacks and blast radius named)
• Dependency correctness (no inverted edges)
• Acceptance clarity (observable PASS/FAIL per step)

Close

WIN: PASS|FAIL
SG: sg-0072 plan-seal
EVIDENCE:
- ...

Constraints & Gotchas

• Plans without file paths = unexecutable theater; seal must FAIL them.
• Missing migration/rollback gates make cutovers unsafe — require both directions.
• Parallel workstreams without interface contracts thrash shared types/APIs.
• Implicit open questions become silent wrong assumptions; surface or close them.
• Over-planning past the first vertical slice delays learning; seal the slice, park the rest.
• ADR without rejected alternatives is a press release, not a decision record.
• Acceptance criteria that only say “works” are not sealable — demand observable checks.
• Do not use outside Planning & architecture (route via /cat-plan or /opgrok).

Anti-patterns

• Implementing production code under a pure plan mandate
• Vague timelines with no dependency edges
• Skipping non-goals or “out of scope”
• Sealing vibe diagrams / whiteboard photos without path mapping
• Freezing a plan that still lists undecided options as co-equal
• Do not write exploits, malware, or undisclosed destructive automation

Definition of Done

• Deliverable matches brief under seal for Planning & architecture.
• Win-gate evidence attached: ordered steps, files/owners, risks/rollbacks, acceptance, non-goals.
• WIN: PASS with concrete evidence paths/commands; FAIL otherwise.
• Downstream SuperGroks consume outputs with zero clarification loops.

Optional Tool Surface

• list_dir, grep -r / grep -l — architecture anchors, ADR index, interface hits
• read_file — existing ADRs, RFCs, DESIGN.md, package manifests
• cargo metadata --no-deps, workspace package graphs — build/dependency order
• SuperGrok registry — hire-plan targets for handoff
• Agent tools: read_file, list_dir, grep
• Binary id: opgrok.sg.plan-seal

References

• core/skills/plan/SKILL.md (category navigator)
• core/tools/domain_enrichment.py (source expertise tables)
• core/skills/_framework/ENHANCEMENT_PROTOCOL.md
• core/skills/_framework/NAVIGATION.md