@bonesofspring/ai-rules 0.2.27 → 0.2.28

package/presets/opencode/java/agents/integration-test-generator.md

@@ -6,6 +6,23 @@ permission:
   edit: allow
 ---
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-integration.md`
+- `.opencode/rules/tests-e2e.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are the **integration-test-generator** for Java hexagonal services.
 
 ## Scope

package/presets/opencode/java/agents/integration-test-healer.md

@@ -6,6 +6,23 @@ permission:
   edit: allow
 ---
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-integration.md`
+- `.opencode/rules/tests-e2e.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are the **integration-test-healer** for Java hexagonal services.
 
 ## Scope

package/presets/opencode/java/agents/integration-test-planner.md

@@ -10,6 +10,23 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-integration.md`
+- `.opencode/rules/tests-e2e.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are the **integration-test-planner** for Java hexagonal services.
 
 ## Scope

package/presets/opencode/java/agents/migration-specialist.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You plan migrations for **Java modules** (toolchain, deps, layout refactors). Do not edit production code.
 
 ## Inputs

package/presets/opencode/java/agents/openspec-author.md

@@ -11,6 +11,21 @@ permission:
 
 Write team artifacts under `.ai.local/team/**` and scoped files (`openspec/**`). Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are an **OpenSpec** (Fission AI spec-driven SDD) author for a frontend/backend agent team.
 
 Your only writes target `openspec/**` (and team artifact manifests). Production code lives behind `feature-developer`; you are **docs / spec only**.

package/presets/opencode/java/agents/performance-auditor.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Audit Java performance risks on the task scope.
 
 ## Focus

package/presets/opencode/java/agents/qa-tester.md

@@ -6,6 +6,24 @@ permission:
   edit: allow
 ---
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-unit.md`
+- `.opencode/rules/tests-integration.md`
+- `.opencode/rules/tests-e2e.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Ensure test coverage for the Java task scope.
 
 ## Prefer

package/presets/opencode/java/agents/security-reviewer.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Review Java diffs against `security-java.md`.
 
 ## Focus

package/presets/opencode/java/agents/solution-architect.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Design solutions for **Java hexagonal** services.
 
 ## Deliverable

package/presets/opencode/java/agents/task-analyst.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Analyze and decompose work for **Java hexagonal** backends.
 
 ## Classify first

package/presets/opencode/java/agents/task-router.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are a task router for a **Java hexagonal backend** agent team. You **do not** implement tasks — you classify intent and plan the agent pipeline.
 
 ## Inputs

package/presets/opencode/java/agents/tech-writer.md

@@ -11,6 +11,21 @@ permission:
 
 Write team artifacts under `.ai.local/team/**` and scoped files (`docs/**`). Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Write documentation from task artifacts (brief, architecture, implementation, review).
 
 ## Java preset / services

package/presets/opencode/java/agents/unit-test-generator.md

@@ -6,6 +6,22 @@ permission:
   edit: allow
 ---
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-unit.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Generate `*Test.java` / `*Tests.java` from the plan. Prefer parameterized JUnit tests and port fakes/mocks (Mockito if present).
 
 ## Rules

package/presets/opencode/java/agents/unit-test-healer.md

@@ -6,6 +6,22 @@ permission:
   edit: allow
 ---
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-unit.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Heal failing `mvn test` / `./gradlew test` unit failures. Prefer fixing tests/fakes; only change production if assertion proves a bug and brief allows.
 
 Follow `tests-unit.md`. Never introduce Playwright/XCUITest.

package/presets/opencode/java/agents/unit-test-planner.md

@@ -10,6 +10,22 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/java-app-core.md`
+- `.opencode/rules/architecture-boundaries.md` — layer boundaries and allowed imports
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/tests-unit.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Plan **Java unit tests** (JUnit 5 `@ParameterizedTest`, AssertJ). No production edits.
 
 ## Focus

package/presets/opencode/java/opencode.json

@@ -1,5 +1,19 @@
 {
   "$schema": "https://opencode.ai/config.json",
+  "instructions": [
+    ".opencode/AGENTS.md",
+    ".opencode/rules/java-app-core.md",
+    ".opencode/rules/architecture-boundaries.md",
+    ".opencode/rules/code-quality-and-refactoring.md",
+    ".opencode/rules/ai-local-overlay.md"
+  ],
+  "permission": {
+    "bash": {
+      "*": "allow",
+      "git commit *": "deny",
+      "git push *": "deny"
+    }
+  },
   "mcp": {
     "context7": {
       "type": "local",

package/presets/opencode/java/rules/README.md

@@ -4,8 +4,13 @@ Plain Markdown. OpenCode does **not** auto-load these files. Root `AGENTS.md`
 and `.opencode/AGENTS.md` list four essential paths to Read immediately; the
 rest are lazy-loaded when the task matches.
 
-Do not put this catalog into `opencode.json` `instructions` as the only
-always-on path (V2 may ignore that field).
+Subagent tasks (`Task`/`@name`) start blank and never read these rule files;
+the **authoritative** mechanism is the `## Rules to load (subagent context)`
+block injected into every `.opencode/agents/*.md`. The preset also adds a
+compact `opencode.json` `instructions` list (alwaysOn paths +
+`.opencode/AGENTS.md`) as a hedge for the parent context, with
+`bin/opencode-install.js` filling-if-absent on re-init; do **not** treat
+`instructions` as the only always-on path (V2 may ignore the field).
 
 ## Essential (entry-critical)
 

package/presets/opencode/mcp-ts/OPENCODE.md

@@ -24,6 +24,19 @@ Root `AGENTS.md` is native always-on. Read it first, then this file.
 
 See `.opencode/rules/README.md` for the full catalog and loading strategy.
 
+## Harness wiring
+
+- `opencode.json` carries a compact `instructions` list (always-on rule
+  paths + `.opencode/AGENTS.md`) and a harness-level `permission.bash`
+  block. On re-init, `mergeOpencodeConfig` unions preset instruction paths
+  into the consumer list and fills missing bash keys. Extra paths and
+  same-key overrides stay.
+- **Subagent rules** are re-declared inside every `.opencode/agents/*.md`
+  via the `## Rules to load (subagent context)` block (always-on + zone
+  rules filtered to stems that exist for this stack). `Task`-launched
+  subagents do not inherit the parent's `instructions`, so this block
+  is the authoritative mechanism.
+
 **Platform docs:** published with the `@bonesofspring/ai-rules` package
 (`docs/OPENCODE.md`, `docs/OPENCODE-COMPATIBILITY.md`) — **not** copied into
 the consumer repo. Source: https://github.com/bonesofspring/ai-rules
@@ -48,4 +61,5 @@ npx @bonesofspring/ai-rules init opencode --preset mcp-ts
 
 Copies rules, commands, agents, hooks, skills, `OPENCODE.md` →
 `.opencode/AGENTS.md`, `MCP.md`, `opencode.json` → project `opencode.json`
-(`mcp` merge), plus `REPO_AGENTS.md` → root `AGENTS.md`. Seeds `.ai.local/team/`.
+(`mcp` merge, instruction-path union, missing bash keys), plus
+`REPO_AGENTS.md` → root `AGENTS.md`. Seeds `.ai.local/team/`.

package/presets/opencode/mcp-ts/adapter-map.json

@@ -183,7 +183,7 @@
       "category": "mcp",
       "disposition": "transform",
       "loadMode": "manual",
-      "reason": "Cursor mcp.json → opencode.json mcp (local/remote types)"
+      "reason": "Cursor mcp.json → opencode.json mcp (local/remote types); emits permission.bash + compact instructions"
     },
     {
       "source": "rules/agent-team-intake.mdc",

package/presets/opencode/mcp-ts/agents/build-verifier.md

@@ -10,6 +10,23 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/mcp-ts-app-core.md`
+- `.opencode/rules/mcp-server-boundaries.md`
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/mcp-ts-tooling.md`
+- `.opencode/rules/post-change-test.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are the build verifier for **TypeScript MCP servers** (and for **ai-rules preset packaging** when scope says so).
 
 ## Consumer MCP repos

package/presets/opencode/mcp-ts/agents/feature-developer.md

@@ -6,6 +6,28 @@ permission:
   edit: allow
 ---
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/mcp-ts-app-core.md`
+- `.opencode/rules/mcp-server-boundaries.md`
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+- `.opencode/rules/mcp-tool-design.md`
+- `.opencode/rules/mcp-contract-testing.md`
+- `.opencode/rules/mcp-auth-secrets.md`
+- `.opencode/rules/mcp-transport-packaging.md`
+- `.opencode/rules/mcp-ts-conventions.md`
+- `.opencode/rules/mcp-ts-tooling.md`
+- `.opencode/rules/post-change-test.md`
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Implement MCP server features following **mcp-ts** boundaries.
 
 ## Mandatory flow

package/presets/opencode/mcp-ts/agents/security-reviewer.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/mcp-ts-app-core.md`
+- `.opencode/rules/mcp-server-boundaries.md`
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Review MCP diffs against `security-mcp.md` and `mcp-auth-secrets.md`.
 
 ## Focus

package/presets/opencode/mcp-ts/agents/solution-architect.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/mcp-ts-app-core.md`
+- `.opencode/rules/mcp-server-boundaries.md`
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Design MCP server changes for **mcp-ts**.
 
 ## Deliverable

package/presets/opencode/mcp-ts/agents/task-analyst.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/mcp-ts-app-core.md`
+- `.opencode/rules/mcp-server-boundaries.md`
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 Analyze MCP server work requests for the **mcp-ts** preset.
 
 ## Classify first

package/presets/opencode/mcp-ts/agents/task-router.md

@@ -10,6 +10,21 @@ permission:
 
 Write only team artifacts under `.ai.local/team/**`. Do not edit production code.
 
+## Rules to load (subagent context)
+
+This subagent does **not** inherit `opencode.json` `instructions`. Read these rules explicitly before acting:
+
+- `.opencode/rules/mcp-ts-app-core.md`
+- `.opencode/rules/mcp-server-boundaries.md`
+- `.opencode/rules/code-quality-and-refactoring.md`
+- `.opencode/rules/ai-local-overlay.md`
+- `.opencode/rules/feature-delivery-workflow.md` — layer order, mockup gate, ux-design gate
+- `.opencode/rules/reference-features.md` — analog files first
+- `.opencode/rules/anti-sycophancy-discipline.md` — procedural anti-sycophancy rules
+- `.opencode/rules/mcp-usage.md` — when verifying third-party APIs (Context7)
+
+Also read matching `.ai.local/rules/*.md` (local overlay) for the active zone, and the project strict rules if present.
+
 You are a task router for a **TypeScript MCP server** agent team. You **do not** implement tasks — you classify intent and plan the agent pipeline.
 
 ## Inputs

package/presets/opencode/mcp-ts/opencode.json

@@ -1,5 +1,19 @@
 {
   "$schema": "https://opencode.ai/config.json",
+  "instructions": [
+    ".opencode/AGENTS.md",
+    ".opencode/rules/mcp-ts-app-core.md",
+    ".opencode/rules/mcp-server-boundaries.md",
+    ".opencode/rules/code-quality-and-refactoring.md",
+    ".opencode/rules/ai-local-overlay.md"
+  ],
+  "permission": {
+    "bash": {
+      "*": "allow",
+      "git commit *": "deny",
+      "git push *": "deny"
+    }
+  },
   "mcp": {
     "context7": {
       "type": "local",

package/presets/opencode/mcp-ts/rules/README.md

@@ -4,8 +4,13 @@ Plain Markdown. OpenCode does **not** auto-load these files. Root `AGENTS.md`
 and `.opencode/AGENTS.md` list four essential paths to Read immediately; the
 rest are lazy-loaded when the task matches.
 
-Do not put this catalog into `opencode.json` `instructions` as the only
-always-on path (V2 may ignore that field).
+Subagent tasks (`Task`/`@name`) start blank and never read these rule files;
+the **authoritative** mechanism is the `## Rules to load (subagent context)`
+block injected into every `.opencode/agents/*.md`. The preset also adds a
+compact `opencode.json` `instructions` list (alwaysOn paths +
+`.opencode/AGENTS.md`) as a hedge for the parent context, with
+`bin/opencode-install.js` filling-if-absent on re-init; do **not** treat
+`instructions` as the only always-on path (V2 may ignore the field).
 
 ## Essential (entry-critical)
 

package/presets/opencode/next/OPENCODE.md

@@ -24,6 +24,19 @@ Root `AGENTS.md` is native always-on. Read it first, then this file.
 
 See `.opencode/rules/README.md` for the full catalog and loading strategy.
 
+## Harness wiring
+
+- `opencode.json` carries a compact `instructions` list (always-on rule
+  paths + `.opencode/AGENTS.md`) and a harness-level `permission.bash`
+  block. On re-init, `mergeOpencodeConfig` unions preset instruction paths
+  into the consumer list and fills missing bash keys. Extra paths and
+  same-key overrides stay.
+- **Subagent rules** are re-declared inside every `.opencode/agents/*.md`
+  via the `## Rules to load (subagent context)` block (always-on + zone
+  rules filtered to stems that exist for this stack). `Task`-launched
+  subagents do not inherit the parent's `instructions`, so this block
+  is the authoritative mechanism.
+
 **Platform docs:** published with the `@bonesofspring/ai-rules` package
 (`docs/OPENCODE.md`, `docs/OPENCODE-COMPATIBILITY.md`) — **not** copied into
 the consumer repo. Source: https://github.com/bonesofspring/ai-rules
@@ -62,4 +75,5 @@ npx @bonesofspring/ai-rules init opencode --preset next
 
 Copies rules, commands, agents, hooks, skills, `OPENCODE.md` →
 `.opencode/AGENTS.md`, `MCP.md`, `opencode.json` → project `opencode.json`
-(`mcp` merge), plus `REPO_AGENTS.md` → root `AGENTS.md`. Seeds `.ai.local/team/`.
+(`mcp` merge, instruction-path union, missing bash keys), plus
+`REPO_AGENTS.md` → root `AGENTS.md`. Seeds `.ai.local/team/`.