nccgs 1.1.0 → 1.3.0

package/.claude/skills/milestone-review/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: milestone-review
 description: Evaluates milestone progress, scope, quality, risk, evidence, and release trajectory against approved goals without disguising incomplete work as percentage completion.
-allowed-tools: Read, Glob, Grep, Bash, Task
+allowed-tools: Read, Glob, Grep, Bash, Agent
 ---
 
 # Milestone Review
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Remain read-only unless asked to update records. Compare approved milestone outcomes with feature closures, evidence, defects, dependencies, PO gates, and repository/build state. Report achieved outcomes, incomplete gates, scope movement, forecast assumptions, top risks, and decision options. Use evidence-based confidence rather than arbitrary completion percentages.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/milestone-review/references/astra.md

@@ -0,0 +1,12 @@
+# Milestone Review — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Astra owns this workflow and its canonical outputs. Claude only inspects relevant
+inputs and returns factual constraints, missing information, or proposed amendments.
+Do not create a replacement plan/document or sign off as Astra.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/milestone-review/references/legacy.md

@@ -0,0 +1,7 @@
+# Milestone Review
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Remain read-only unless asked to update records. Compare approved milestone outcomes with feature closures, evidence, defects, dependencies, PO gates, and repository/build state. Report achieved outcomes, incomplete gates, scope movement, forecast assumptions, top risks, and decision options. Use evidence-based confidence rather than arbitrary completion percentages.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/nccgs-code-review/SKILL.md

@@ -0,0 +1,15 @@
+---
+name: nccgs-code-review
+description: Performs a read-only code review against requirements, architecture, Unity constraints, ownership, tests, performance, security, and regression risk.
+allowed-tools: Read, Glob, Grep, Bash, Agent
+---
+
+# Code Review
+
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
+
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
+
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/nccgs-code-review/references/astra.md

@@ -0,0 +1,15 @@
+# Code Review — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Run the independent reviewer selected by pipeline.json on the final snapshot.
+Read raw requirements, diff, affected callers, and verification evidence. Return
+PASS, CHANGES_REQUIRED, or BLOCKED with findings, severity, locations, and evidence.
+Do not fix code. BLOCKER/HIGH findings and unmet required criteria prevent PASS.
+Record the review through task record; return fixes to the implementer and review
+again after changes. Review is mandatory even for FAST work.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/{code-review/SKILL.md → nccgs-code-review/references/legacy.md}

@@ -1,9 +1,3 @@
----
-name: code-review
-description: Performs a read-only code review against requirements, architecture, Unity constraints, ownership, tests, performance, security, and regression risk.
-allowed-tools: Read, Glob, Grep, Bash, Task
----
-
 # Code Review
 
 Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
@@ -11,4 +5,3 @@ Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet whe
 Read the context packet, final diff, governing decisions, affected callers, and tests. Route distinct Unity, networking, security, performance, or UI lenses when needed. Report only actionable findings with severity, exact evidence, impact, and remediation direction. Verify that tests assert requirements and that serialized/lifecycle behavior is addressed. Do not edit or praise generally; state when no material issue is found.
 
 Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
-

package/.claude/skills/performance-audit/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: performance-audit
 description: Profiles or reviews Unity CPU, GPU, memory, loading, allocations, and frame pacing against project budgets and reproducible scenarios.
-allowed-tools: Read, Glob, Grep, Bash, Task
+allowed-tools: Read, Glob, Grep, Bash, Agent
 ---
 
 # Performance Audit
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Remain read-only unless explicitly asked to optimize. Define platform, scene, hardware/environment, scenario, baseline, and budget. Use official Unity Skills and profiling tools where available. Measure before attributing cause. Report bottlenecks with evidence, player impact, confidence, and prioritized experiment or remediation. For implemented optimizations, compare identical before/after conditions and regression risk.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/performance-audit/references/astra.md

@@ -0,0 +1,17 @@
+# Performance Audit — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Domain objective: Profiles or reviews Unity CPU, GPU, memory, loading, allocations, and frame pacing against project budgets and reproducible scenarios.
+
+Work only inside the supplied plan and assigned write set. Activate an optional
+specialist with the agents command only when its domain is required. Keep one writer
+per path and use official Unity Skills for Editor/assets/runtime operations.
+
+This procedure is read-only: return evidence-backed findings and decisions to Astra
+or the implementer. Supplemental analysis does not replace the mandatory Fable gate.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/performance-audit/references/legacy.md

@@ -0,0 +1,7 @@
+# Performance Audit
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Remain read-only unless explicitly asked to optimize. Define platform, scene, hardware/environment, scenario, baseline, and budget. Use official Unity Skills and profiling tools where available. Measure before attributing cause. Report bottlenecks with evidence, player impact, confidence, and prioritized experiment or remediation. For implemented optimizations, compare identical before/after conditions and regression risk.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/plan-feature/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: plan-feature
 description: Creates an executable feature plan with dependencies, role ownership, write sets, verification, decisions, and closure gates. Use for meaningful STANDARD or CONTROLLED implementation before edits begin.
-allowed-tools: Read, Glob, Grep, Write, Edit, Task, AskUserQuestion
+allowed-tools: Read, Glob, Grep, Write, Edit, Agent, AskUserQuestion
 ---
 
 # Plan Feature
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Read policy, state, feature contract, and context packet. Define approved outcome, non-goals, risk, execution DAG, role per node, one writer per path, checkpoints, rollback, evidence, PO decisions, and closure artifacts. Use the [orchestration protocol](../../nccgs/protocols/orchestration.md). Parallelize only independent nodes. Persist the plan in the mapped planning source or `.nccgs/features/` when no canonical plan exists. A plan authorizes no extra product scope.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/plan-feature/references/astra.md

@@ -0,0 +1,12 @@
+# Plan Feature — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Astra owns this workflow and its canonical outputs. Claude only inspects relevant
+inputs and returns factual constraints, missing information, or proposed amendments.
+Do not create a replacement plan/document or sign off as Astra.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/plan-feature/references/legacy.md

@@ -0,0 +1,7 @@
+# Plan Feature
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Read policy, state, feature contract, and context packet. Define approved outcome, non-goals, risk, execution DAG, role per node, one writer per path, checkpoints, rollback, evidence, PO decisions, and closure artifacts. Use the [orchestration protocol](../../../nccgs/protocols/orchestration.md). Parallelize only independent nodes. Persist the plan in the mapped planning source or `.nccgs/features/` when no canonical plan exists. A plan authorizes no extra product scope.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/playtest/SKILL.md

@@ -6,9 +6,10 @@ allowed-tools: Read, Glob, Grep, Write, Edit, AskUserQuestion
 
 # Playtest
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Define build/version, participant assumptions, scenario, setup, tasks, observation prompts, success signals, and questions without leading the participant. Keep objective telemetry separate from subjective feedback. Record observations and decisions in the mapped playtest source or `.nccgs/playtests/`. A model may prepare and summarize; it must not fabricate player experience or close the PO gate.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/playtest/references/astra.md

@@ -0,0 +1,20 @@
+# Playtest — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Domain objective: Prepares and records a reproducible Product Owner or player playtest for subjective experience, usability, pacing, balance, readability, audio, or narrative gates.
+
+Work only inside the supplied plan and assigned write set. Activate an optional
+specialist with the agents command only when its domain is required. Keep one writer
+per path and use official Unity Skills for Editor/assets/runtime operations.
+
+For code changes: task init with Astra's plan, bind the session, implement and test,
+record implementation evidence, request independent review, fix and re-review as
+needed, then task check --target report. Only after that check passes write and
+record the report for Astra. Stop at AWAITING_ASTRA_CHECK. Astra checking and the
+Product Owner's explicit acceptance happen outside the Claude implementation session.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/playtest/references/legacy.md

@@ -0,0 +1,7 @@
+# Playtest
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Define build/version, participant assumptions, scenario, setup, tasks, observation prompts, success signals, and questions without leading the participant. Keep objective telemetry separate from subjective feedback. Record observations and decisions in the mapped playtest source or `.nccgs/playtests/`. A model may prepare and summarize; it must not fabricate player experience or close the PO gate.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/project-stage/SKILL.md

@@ -6,9 +6,10 @@ allowed-tools: Read, Glob, Grep, Bash
 
 # Project Stage
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Remain read-only. Inspect policy, state, canonical sources, code/assets, tests, builds, release evidence, and Git history. Classify discovery, concept, pre-production, production, stabilization, release, live operations, maintenance, or mixed/transition state. Report evidence, missing prerequisites that materially block the next outcome, and one recommended next workflow. Existing projects need not backfill ceremonial phases.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/project-stage/references/astra.md

@@ -0,0 +1,12 @@
+# Project Stage — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Astra owns this workflow and its canonical outputs. Claude only inspects relevant
+inputs and returns factual constraints, missing information, or proposed amendments.
+Do not create a replacement plan/document or sign off as Astra.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/project-stage/references/legacy.md

@@ -0,0 +1,7 @@
+# Project Stage
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Remain read-only. Inspect policy, state, canonical sources, code/assets, tests, builds, release evidence, and Git history. Classify discovery, concept, pre-production, production, stabilization, release, live operations, maintenance, or mixed/transition state. Report evidence, missing prerequisites that materially block the next outcome, and one recommended next workflow. Existing projects need not backfill ceremonial phases.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/prototype-feature/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: prototype-feature
 description: Creates an isolated, time-boxed Unity prototype to answer one explicit design or technical hypothesis, with measurement and keep/discard criteria.
-allowed-tools: Read, Glob, Grep, Write, Edit, Bash, Task, AskUserQuestion
+allowed-tools: Read, Glob, Grep, Write, Edit, Bash, Agent, AskUserQuestion
 ---
 
 # Prototype Feature
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Define hypothesis, cheapest falsifiable experiment, isolation boundary, timebox, measurement, success/failure threshold, and disposal plan. Keep production contracts and source untouched unless integration is explicitly approved. Use official Unity Skills for operations. Record result, evidence, limitations, and recommendation: discard, iterate, or migrate through a separate CONTROLLED plan. Never silently promote prototype code.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/prototype-feature/references/astra.md

@@ -0,0 +1,20 @@
+# Prototype Feature — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Domain objective: Creates an isolated, time-boxed Unity prototype to answer one explicit design or technical hypothesis, with measurement and keep/discard criteria.
+
+Work only inside the supplied plan and assigned write set. Activate an optional
+specialist with the agents command only when its domain is required. Keep one writer
+per path and use official Unity Skills for Editor/assets/runtime operations.
+
+For code changes: task init with Astra's plan, bind the session, implement and test,
+record implementation evidence, request independent review, fix and re-review as
+needed, then task check --target report. Only after that check passes write and
+record the report for Astra. Stop at AWAITING_ASTRA_CHECK. Astra checking and the
+Product Owner's explicit acceptance happen outside the Claude implementation session.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/prototype-feature/references/legacy.md

@@ -0,0 +1,7 @@
+# Prototype Feature
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Define hypothesis, cheapest falsifiable experiment, isolation boundary, timebox, measurement, success/failure threshold, and disposal plan. Keep production contracts and source untouched unless integration is explicitly approved. Use official Unity Skills for operations. Record result, evidence, limitations, and recommendation: discard, iterate, or migrate through a separate CONTROLLED plan. Never silently promote prototype code.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/qa-plan/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: qa-plan
 description: Builds a risk-based Unity QA plan mapping requirements to automated, integration, runtime, build, performance, compatibility, and Product Owner validation scenarios.
-allowed-tools: Read, Glob, Grep, Write, Edit, Task
+allowed-tools: Read, Glob, Grep, Write, Edit, Agent
 ---
 
 # Qa Plan
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Read policy, context packet, requirements, architecture, affected systems, existing tests, and baseline. Identify risk surfaces and map each criterion to the strongest practical evidence, test owner, environment, data/fixtures, expected result, regression scope, and PO gate. Include failure triage and stopping conditions. Persist in the mapped QA source or `.nccgs/evidence/`. Do not inflate coverage with redundant cases.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/qa-plan/references/astra.md

@@ -0,0 +1,12 @@
+# Qa Plan — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Astra owns this workflow and its canonical outputs. Claude only inspects relevant
+inputs and returns factual constraints, missing information, or proposed amendments.
+Do not create a replacement plan/document or sign off as Astra.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/qa-plan/references/legacy.md

@@ -0,0 +1,7 @@
+# Qa Plan
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Read policy, context packet, requirements, architecture, affected systems, existing tests, and baseline. Identify risk surfaces and map each criterion to the strongest practical evidence, test owner, environment, data/fixtures, expected result, regression scope, and PO gate. Include failure triage and stopping conditions. Persist in the mapped QA source or `.nccgs/evidence/`. Do not inflate coverage with redundant cases.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/release/SKILL.md

@@ -1,18 +1,15 @@
 ---
 name: release
 description: Prepares a reproducible Unity release candidate and coordinates builds, versions, regression, known issues, rollback, and sign-offs. Publishing, deployment, store submission, or production changes require explicit Product Owner authority.
-allowed-tools: Read, Glob, Grep, Write, Edit, Bash, Task, AskUserQuestion, WebSearch, WebFetch
+allowed-tools: Read, Glob, Grep, Write, Edit, Bash, Agent, AskUserQuestion, WebSearch, WebFetch
 ---
 
 # Release
 
-Treat release work as CONTROLLED when it changes platform configuration, signing, dependencies/services, compatibility, or production state.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Read release policy, candidate scope, closures, known issues, migrations, versioning, repository state, build matrix, and rollback requirements. Run `/release-readiness` before producing a candidate.
-
-Compose release lead, release engineer, Unity build specialist, QA, performance, security, localization, and content roles only where the target requires them. Use official Unity Skills for build operations.
-
-Produce local artifacts, logs, manifests, checksums, evidence, and rollback notes. Fix authorized technical failures but never waive gates or change scope silently. Immediately before upload, publication, store submission, deployment, signing changes, or production configuration, obtain explicit Product Owner authority.
-
-Report candidate identity, source revision, targets, tests/builds, artifact paths, known issues, waivers, pending gates, rollback, and external actions intentionally not performed.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/release/references/astra.md

@@ -0,0 +1,23 @@
+# Release — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Domain objective: Prepares a reproducible Unity release candidate and coordinates builds, versions, regression, known issues, rollback, and sign-offs. Publishing, deployment, store submission, or production changes require explicit Product Owner authority.
+
+Work only inside the supplied plan and assigned write set. Activate an optional
+specialist with the agents command only when its domain is required. Keep one writer
+per path and use official Unity Skills for Editor/assets/runtime operations.
+
+For code changes: task init with Astra's plan, bind the session, implement and test,
+record implementation evidence, request independent review, fix and re-review as
+needed, then task check --target report. Only after that check passes write and
+record the report for Astra. Stop at AWAITING_ASTRA_CHECK. Astra checking and the
+Product Owner's explicit acceptance happen outside the Claude implementation session.
+
+Publication and other external release actions still require explicit Product Owner
+authority; successful local verification grants no publication permission.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/release/references/legacy.md

@@ -0,0 +1,11 @@
+# Release
+
+Treat release work as CONTROLLED when it changes platform configuration, signing, dependencies/services, compatibility, or production state.
+
+Read release policy, candidate scope, closures, known issues, migrations, versioning, repository state, build matrix, and rollback requirements. Run `/release-readiness` before producing a candidate.
+
+Compose release lead, release engineer, Unity build specialist, QA, performance, security, localization, and content roles only where the target requires them. Use official Unity Skills for build operations.
+
+Produce local artifacts, logs, manifests, checksums, evidence, and rollback notes. Fix authorized technical failures but never waive gates or change scope silently. Immediately before upload, publication, store submission, deployment, signing changes, or production configuration, obtain explicit Product Owner authority.
+
+Report candidate identity, source revision, targets, tests/builds, artifact paths, known issues, waivers, pending gates, rollback, and external actions intentionally not performed.

package/.claude/skills/release-readiness/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: release-readiness
 description: Performs a read-only go/no-go readiness review across feature closure, tests, builds, compatibility, performance, security, localization, accessibility, known issues, rollback, and Product Owner sign-offs.
-allowed-tools: Read, Glob, Grep, Bash, Task
+allowed-tools: Read, Glob, Grep, Bash, Agent
 ---
 
 # Release Readiness
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Read release policy, candidate identity, closures, evidence, defects, build matrix, migrations, and rollback plan. Route only applicable domain checks. Classify every gate PASS, FAIL, NOT PROVEN, PO GATE, or waived with authority. Return GO, CONDITIONAL GO, or NO-GO with exact blockers and residual risks. Do not build, publish, or alter release state.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/release-readiness/references/astra.md

@@ -0,0 +1,17 @@
+# Release Readiness — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Domain objective: Performs a read-only go/no-go readiness review across feature closure, tests, builds, compatibility, performance, security, localization, accessibility, known issues, rollback, and Product Owner sign-offs.
+
+Work only inside the supplied plan and assigned write set. Activate an optional
+specialist with the agents command only when its domain is required. Keep one writer
+per path and use official Unity Skills for Editor/assets/runtime operations.
+
+This procedure is read-only: return evidence-backed findings and decisions to Astra
+or the implementer. Supplemental analysis does not replace the mandatory Fable gate.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/release-readiness/references/legacy.md

@@ -0,0 +1,7 @@
+# Release Readiness
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Read release policy, candidate identity, closures, evidence, defects, build matrix, migrations, and rollback plan. Route only applicable domain checks. Classify every gate PASS, FAIL, NOT PROVEN, PO GATE, or waived with authority. Return GO, CONDITIONAL GO, or NO-GO with exact blockers and residual risks. Do not build, publish, or alter release state.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/retrospective/SKILL.md

@@ -1,14 +1,15 @@
 ---
 name: retrospective
 description: Captures evidence-based process improvements after a feature, sprint, release, migration, or incident without turning one anecdote into universal policy.
-allowed-tools: Read, Glob, Grep, Write, Edit, Task
+allowed-tools: Read, Glob, Grep, Write, Edit, Agent
 ---
 
 # Retrospective
 
-Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Read goals, state transitions, handoffs, evidence, blockers, rework, and outcomes. Identify what helped, what failed, causal evidence, and a small number of testable improvements with owner and review trigger. Separate project-specific learning from framework change. Update canonical process records only when the improvement is accepted; do not add permanent rules for unverified anecdotes.
-
-Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/retrospective/references/astra.md

@@ -0,0 +1,12 @@
+# Retrospective — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Astra owns this workflow and its canonical outputs. Claude only inspects relevant
+inputs and returns factual constraints, missing information, or proposed amendments.
+Do not create a replacement plan/document or sign off as Astra.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/retrospective/references/legacy.md

@@ -0,0 +1,7 @@
+# Retrospective
+
+Read `.nccgs/project.yaml`, `.nccgs/state.md`, and the active context packet when present. Preserve Product Owner authority and unrelated work.
+
+Read goals, state transitions, handoffs, evidence, blockers, rework, and outcomes. Identify what helped, what failed, causal evidence, and a small number of testable improvements with owner and review trigger. Separate project-specific learning from framework change. Update canonical process records only when the improvement is accepted; do not add permanent rules for unverified anecdotes.
+
+Return outcome, evidence, affected records, blockers or decisions, residual risk, and the exact next workflow. Do not claim DONE unless this is the `closure` skill and every configured gate passes.

package/.claude/skills/review/SKILL.md

@@ -1,16 +1,15 @@
 ---
 name: review
 description: Runs an independent, adversarial, read-only review of code, design, architecture, migration, release, or feature evidence. Use for second opinions and pre-closure scrutiny; it does not implement fixes.
-allowed-tools: Read, Glob, Grep, Bash, Task
+allowed-tools: Read, Glob, Grep, Bash, Agent
 ---
 
 # Review
 
-Remain read-only. Review from raw requirements, policy, governing decisions, final diff, tests, and runtime/build evidence rather than implementer rationale.
+Read models.profile in .nccgs/project.yaml; if absent, use the defaultProfile in
+.claude/nccgs/pipeline.json. Load exactly one procedure:
 
-Select domain reviewers only when they add a distinct failure lens. CONTROLLED work must include `nccgs-adversarial-reviewer`; provide bounded artifacts and no intended verdict.
-
-Check behavior, edge cases, design/canon, architecture, Unity lifecycle/serialization, ownership, compatibility, security, performance, test coverage, migration reversibility, and closure claims as applicable.
-
-Each finding includes BLOCKER/HIGH/MEDIUM/LOW severity, location/evidence, violated contract or reproducible failure mode, impact, and smallest safe remediation direction. Report unknowns and confidence. Lead with findings; do not edit or drift into fixes.
+- astra-claude: [Astra/Claude procedure](references/astra.md).
+- quality, balanced, inherit: [Legacy procedure](references/legacy.md).
 
+Do not load or merge the other branch. Unknown profiles are configuration errors.

package/.claude/skills/review/references/astra.md

@@ -0,0 +1,15 @@
+# Review — astra-claude
+
+Read [the pipeline protocol](../../../nccgs/protocols/astra-claude.md) and
+[generated routing](../../../nccgs/routing.generated.md). Use Astra's supplied plan,
+requirements, and exact context references. Never substitute a legacy procedure.
+
+Run the independent reviewer selected by pipeline.json on the final snapshot.
+Read raw requirements, diff, affected callers, and verification evidence. Return
+PASS, CHANGES_REQUIRED, or BLOCKED with findings, severity, locations, and evidence.
+Do not fix code. BLOCKER/HIGH findings and unmet required criteria prevent PASS.
+Record the review through task record; return fixes to the implementer and review
+again after changes. Review is mandatory even for FAST work.
+
+Use [task gate commands](../../../nccgs/protocols/task-gates.md) for evidence and
+state transitions. Report outcome, evidence, pending decisions, and next owner.

package/.claude/skills/review/references/legacy.md

@@ -0,0 +1,9 @@
+# Review
+
+Remain read-only. Review from raw requirements, policy, governing decisions, final diff, tests, and runtime/build evidence rather than implementer rationale.
+
+Select domain reviewers only when they add a distinct failure lens. CONTROLLED work must include `nccgs-adversarial-reviewer`; provide bounded artifacts and no intended verdict.
+
+Check behavior, edge cases, design/canon, architecture, Unity lifecycle/serialization, ownership, compatibility, security, performance, test coverage, migration reversibility, and closure claims as applicable.
+
+Each finding includes BLOCKER/HIGH/MEDIUM/LOW severity, location/evidence, violated contract or reproducible failure mode, impact, and smallest safe remediation direction. Report unknowns and confidence. Lead with findings; do not edit or drift into fixes.