feat(skills): add Summit operational ritual skill pack

skills/summit/README.md

@@ -0,0 +1,13 @@
+# Summit Skill Pack
+
+This pack encodes repeatable Summit operating rituals into portable skills.
+
+## Included skills
+- `atomic-pr-planner`
+- `ga-gate-reviewer`
+- `pilot-proof-pack-writer`
+- `closed-pr-rehydrator`
+- `trust-claim-auditor`
+
+Each skill is intentionally compact: trigger-rich `description` frontmatter and
+strict output contracts in the body.

skills/summit/atomic-pr-planner/SKILL.md

@@ -0,0 +1,42 @@
+---
+name: atomic-pr-planner
+version: 1.0.0
+description: |
+  Breaks a requested Summit feature or initiative into a strict atomic PR stack
+  with acceptance criteria, file targets, risks, tests, and merge order.
+  Use when planning implementation, recovering roadmap execution, or converting
+  strategy into Codex-ready PRs.
+  Triggers on: "make a PR stack", "plan the next PRs", "turn this into atomic
+  PRs", "what are the exact next PRs", "split this into mergeable units".
+---
+
+# atomic-pr-planner
+
+Generate first. Do not ask clarifying questions before producing the first draft.
+
+## Required output
+For each PR provide:
+1. PR title (Conventional Commit style)
+2. Objective
+3. Exact files likely touched
+4. Implementation steps
+5. Tests / validation commands
+6. Acceptance criteria
+7. Risks / rollback plan
+8. Why this PR is atomic
+9. Merge-order dependency
+
+## Quality rules
+- Prefer the smallest mergeable unit.
+- Never mix infra, product, and docs unless required for one testable outcome.
+- Call out hidden coupling explicitly.
+- Fail closed: if a step cannot be validated, mark it **BLOCKED** with reason.
+- Include a "deferred" list for intentionally constrained scope.
+
+## Output format
+Return:
+- `PR-STACK-SUMMARY`
+- `PR-1..N` sections with the nine required fields
+- `DEPENDENCY-DAG` as `PR-x -> PR-y`
+- `BLOCKERS` (or `none`)
+- `DEFERRED` (or `none`)

skills/summit/closed-pr-rehydrator/SKILL.md

@@ -0,0 +1,36 @@
+---
+name: closed-pr-rehydrator
+version: 1.0.0
+description: |
+  Reviews merged or closed PR context and extracts salvageable work into a new
+  merge plan with conflict notes, dependency mapping, and reimplementation
+  sequence. Use when reviving abandoned work, reconciling duplicate PRs, or
+  recovering roadmap execution from stale branches.
+  Triggers on: "rehydrate closed PR", "salvage from old PR", "recover merged
+  context", "duplicate PR cleanup", "extract reusable changes".
+---
+
+# closed-pr-rehydrator
+
+Generate first. Clarify only after presenting the first salvage map.
+
+## Required workflow
+1. Identify source PR scope and intent
+2. Classify content: reusable, obsolete, conflicting, superseded
+3. Produce extraction plan by atomic commit slices
+4. Define validation per slice
+5. Emit merge strategy and rollback path
+
+## Rules
+- Prefer reuse of validated code over rewrite.
+- Explicitly label superseded work and canonical replacement.
+- Surface hidden dependencies before proposing cherry-picks.
+- If provenance is unclear, mark as `QUARANTINED`.
+
+## Output format
+Return sections exactly:
+- `SOURCE-CONTEXT`
+- `SALVAGE-MAP`
+- `ATOMIC-EXTRACTION-PLAN`
+- `VALIDATION-CHECKLIST`
+- `MERGE-AND-ROLLBACK-STRATEGY`

skills/summit/ga-gate-reviewer/SKILL.md

@@ -0,0 +1,41 @@
+---
+name: ga-gate-reviewer
+version: 1.0.0
+description: |
+  Evaluates a Summit branch against GA readiness gates and outputs deterministic
+  PASS/FAIL status with missing evidence and remediation steps.
+  Use when asked to run a GA readiness check, pre-merge gate review, cut-line
+  validation, or release hardening review.
+  Triggers on: "ga gate", "ga readiness", "cut line", "release gates",
+  "what evidence is missing", "is this branch GA-ready".
+---
+
+# ga-gate-reviewer
+
+Generate first. Ask clarifying questions only after the first pass if evidence is missing.
+
+## Inputs to inspect
+- Changed files and affected surfaces
+- Required checks and workflow outcomes
+- Evidence artifacts referenced by governance docs
+- Test, lint, typecheck, and smoke outputs
+
+## Required output
+1. Overall decision: `PASS`, `FAIL`, or `CONDITIONAL-PASS`
+2. Gate matrix table: gate, expected evidence, observed evidence, status
+3. Missing evidence list with exact artifact names/paths
+4. Remediation plan with ordered actions
+5. Residual risk statement with rollback triggers
+
+## Decision policy
+- Mark `FAIL` if any required gate lacks evidence.
+- Mark `CONDITIONAL-PASS` only when risk is bounded and remediation is explicit.
+- Never infer a pass from intent; require traceable artifacts.
+
+## Output format
+Return sections exactly:
+- `GA-DECISION`
+- `GATE-MATRIX`
+- `MISSING-EVIDENCE`
+- `REMEDIATION-PLAN`
+- `RESIDUAL-RISK`

skills/summit/pilot-proof-pack-writer/SKILL.md

@@ -0,0 +1,36 @@
+---
+name: pilot-proof-pack-writer
+version: 1.0.0
+description: |
+  Produces pilot-proof collateral for Summit in a fixed bundle structure:
+  ADR summary, demo narrative, evidence index, validation plan, and rollback
+  notes. Use when preparing pilot readiness packets, customer proof packages,
+  or governance-ready release narratives.
+  Triggers on: "pilot pack", "proof pack", "write ADR/demo/evidence bundle",
+  "pilot readiness artifacts", "customer validation packet".
+---
+
+# pilot-proof-pack-writer
+
+Generate first. Do not delay initial output for clarification.
+
+## Required bundle
+1. ADR snapshot (decision, context, alternatives, consequences)
+2. Demo script (setup, flow, expected outputs, failure handling)
+3. Evidence index (artifact name, source path, verification command)
+4. Validation matrix (requirement -> test -> result placeholder)
+5. Rollback and accountability window
+
+## Constraints
+- Use deterministic headings.
+- Prefer verifiable claims with explicit artifact pointers.
+- If evidence is absent, mark `EVIDENCE-MISSING` and continue.
+- Keep language execution-oriented, not promotional.
+
+## Output format
+Return sections exactly:
+- `ADR-SNAPSHOT`
+- `DEMO-SCRIPT`
+- `EVIDENCE-INDEX`
+- `VALIDATION-MATRIX`
+- `ROLLBACK-AND-MONITORING`

skills/summit/trust-claim-auditor/SKILL.md

@@ -0,0 +1,36 @@
+---
+name: trust-claim-auditor
+version: 1.0.0
+description: |
+  Audits Summit product, compliance, or sales claims against traceable evidence
+  and outputs claim verdicts with confidence, gaps, and remediation actions.
+  Use when validating external-facing assertions, trust-center language,
+  security/compliance statements, or launch messaging.
+  Triggers on: "audit this claim", "is this claim defensible", "trust claim",
+  "evidence-backed messaging", "validate sales statement".
+---
+
+# trust-claim-auditor
+
+Generate first. Do not block on clarifications before initial verdicts.
+
+## Required output per claim
+1. Claim text
+2. Verdict: `SUPPORTED`, `PARTIAL`, `UNSUPPORTED`, or `CONTRADICTED`
+3. Evidence map with exact file/artifact references
+4. Confidence score (0.00-1.00) and basis
+5. Gap list and remediation actions
+
+## Rules
+- No evidence, no support.
+- Distinguish implementation evidence from aspirational roadmap items.
+- Flag legal/compliance-sensitive language for human review.
+- If evidence conflicts, default to `CONTRADICTED` until reconciled.
+
+## Output format
+Return sections exactly:
+- `CLAIM-INVENTORY`
+- `VERDICT-TABLE`
+- `EVIDENCE-MAP`
+- `GAP-REMEDIATION-PLAN`
+- `HUMAN-REVIEW-FLAGS`