chore: sync public export

.github/copilot-instructions.md

@@ -0,0 +1,14 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+# GitHub Copilot Instructions
+
+This repository uses **Kernel** for repo-local agent quality artifacts.
+
+Project: kernel-skills
+
+Before non-trivial changes, inspect `.agent/kernel.yaml` and `.agent/state/current-task.md`.
+Record verification in `.agent/evidence/` before claiming completion.
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/instructions/review.instructions.md

@@ -0,0 +1,15 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+# Kernel Review Instructions
+
+Review agent changes in kernel-skills against Kernel artifacts.
+
+Prime directive: No contract, no implementation. No evidence, no completion. No handoff, no continuity.
+
+- Confirm the task contract matches the diff.
+- Confirm verification evidence supports completion claims.
+- Focus on high-risk paths, generated outputs, and missing tests.
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/instructions/testing.instructions.md

@@ -0,0 +1,13 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+# Kernel Testing Instructions
+
+Use Kernel verification evidence for kernel-skills.
+
+- Prefer targeted tests for local behavior changes.
+- Run project-level checks for refactors and broad changes.
+- Record skipped checks and remaining risks in `.agent/evidence/`.
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/accessibility-scout/SKILL.md

@@ -0,0 +1,47 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: accessibility-scout
+description: "Use for UI changes. Reviews keyboard, focus, semantic HTML, labels, contrast, and ARIA. Do not trigger it for unrelated work just because the skill exists."
+---
+# accessibility-scout
+
+## Purpose
+
+Use for UI changes. Reviews keyboard, focus, semantic HTML, labels, contrast, and ARIA.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Check keyboard path.
+2. Check focus states.
+3. Check labels.
+4. Avoid ARIA misuse.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/adapter-compiler/SKILL.md

@@ -0,0 +1,47 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: adapter-compiler
+description: "Use when generating ADE-specific files from canonical Kernel source. Do not trigger it for unrelated work just because the skill exists."
+---
+# adapter-compiler
+
+## Purpose
+
+Use when generating ADE-specific files from canonical Kernel source.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Read `.agent/kernel.yaml`.
+2. Generate adapter outputs.
+3. Preserve manual sections.
+4. Validate freshness.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/adr-capture/SKILL.md

@@ -0,0 +1,48 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: adr-capture
+description: "Use when an architectural decision or durable tradeoff is made. Do not trigger it for unrelated work just because the skill exists."
+---
+# adr-capture
+
+## Purpose
+
+Use when an architectural decision or durable tradeoff is made.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Record context.
+2. Record decision.
+3. Record alternatives.
+4. Record consequences.
+5. Record follow-up.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/agent-scheduler/SKILL.md

@@ -0,0 +1,47 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: agent-scheduler
+description: "Use when a plan has multiple tasks. Decides serial vs parallel execution. Do not trigger it for unrelated work just because the skill exists."
+---
+# agent-scheduler
+
+## Purpose
+
+Use when a plan has multiple tasks. Decides serial vs parallel execution.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Classify dependencies.
+2. Identify path collisions.
+3. Split safe tasks.
+4. Sequence risky tasks.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/api-contract-guardian/SKILL.md

@@ -0,0 +1,47 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: api-contract-guardian
+description: "Use when changing APIs, SDKs, CLIs, config, schemas, events, migrations, or public types. Do not trigger it for unrelated work just because the skill exists."
+---
+# api-contract-guardian
+
+## Purpose
+
+Use when changing APIs, SDKs, CLIs, config, schemas, events, migrations, or public types.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Detect public contract changes.
+2. Check backward compatibility.
+3. Require docs/changelog updates.
+4. Require contract tests.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/context-router/SKILL.md

@@ -0,0 +1,48 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: context-router
+description: "Use before planning or coding. Builds a curated context bundle from repo maps, changed files, tests, docs, imports, and task contract. Do not trigger it for unrelated work just because the skill exists."
+---
+# context-router
+
+## Purpose
+
+Use before planning or coding. Builds a curated context bundle from repo maps, changed files, tests, docs, imports, and task contract.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Identify relevant files.
+2. Identify relevant tests and commands.
+3. Identify docs and constraints.
+4. Explicitly exclude likely irrelevant context.
+5. Record bundle in `.agent/state/context-bundle.md`.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->

.github/skills/debug-probe/SKILL.md

@@ -0,0 +1,48 @@
+<!-- Generated by Kernel. Edit canonical source under .agent/ instead. -->
+
+---
+name: debug-probe
+description: "Use for bugs, flaky tests, incidents, failing commands, or unclear behavior. Establishes causality before fixes. Do not trigger it for unrelated work just because the skill exists."
+---
+# debug-probe
+
+## Purpose
+
+Use for bugs, flaky tests, incidents, failing commands, or unclear behavior. Establishes causality before fixes.
+
+## Trigger
+
+Use this skill when its purpose matches the current task. Do not trigger it for unrelated work just because the skill exists.
+
+## Workflow
+
+1. Reproduce symptom.
+2. Create causal graph.
+3. Localize boundary.
+4. Add probe.
+5. Avoid fixes before evidence.
+
+## Output
+
+Write a concise artifact under `.agent/` when filesystem access is available. Otherwise, include the artifact in the agent response.
+
+## Quality bar
+
+- Be specific.
+- Prefer observable facts over impressions.
+- Record uncertainty explicitly.
+- Avoid bloating global instruction files.
+- Link output to the current task contract when applicable.
+
+## Related skills
+
+- [[kernel-core]]
+- [[verify-lattice]]
+- [[evidence-ledger]]
+- [[handoff-packet]]
+
+## Manual notes
+
+<!-- kernel:manual:start -->
+
+<!-- kernel:manual:end -->