Refactor ioc

.github/agents/DevOps.agent.md

@@ -2,6 +2,7 @@
 description: "Use when: managing CI/CD pipelines, GitHub Actions workflows, build/test/pack/publish automation, NuGet Trusted Publishing, or release processes. Handles .github/workflows/ files and DevOps configuration."
 model: GPT-5.4 (copilot)
 tools: [vscode/askQuestions, vscode/memory, vscode/resolveMemoryFileUri, execute/getTerminalOutput, execute/runInTerminal, read, agent, edit, search, web, github/get_copilot_job_status, github/get_file_contents, github/get_latest_release, github/get_release_by_tag, github/get_tag, github/issue_read, github/list_branches, github/list_releases, github/list_tags, github/pull_request_read, github/search_code, github/search_issues, github/search_pull_requests, github/search_repositories, 'io.github.upstash/context7/*', 'microsoftdocs/mcp/*', todo, github.vscode-pull-request-github/notification_fetch]
+target: vscode
 agents: ["Explore"]
 user-invocable: true
 argument-hint: "Describe the CI/CD change: add workflow, fix pipeline, update publish config, etc."
@@ -43,7 +44,7 @@ git push origin main && git push origin ioc-v0.9.1-alpha           # triggers pu
 
 **Post-release**: bump `version.json` via `nbgv prepare-release --project <path>` or manual edit.
 
-Follow the **parent agent protocol** in `.github/instructions/memory-policy.instructions.md`.
+Follow the **Parent Workflow** in `.github/instructions/memory-policy.instructions.md`.
 
 ## Approach
 
@@ -78,6 +79,7 @@ Follow the **parent agent protocol** in `.github/instructions/memory-policy.inst
 - 🚫 **Never do:**
 	- Use long-lived API keys (use OIDC temporary credentials only)
 	- Remove tag-based publish gate, `id-token: write`, `NuGet/login@v1`, or `environment: nuget-publish` from publish jobs
+	- Read or write any `/memories/session/*` path with a tool other than #tool:vscode/memory (no #tool:read, #tool:edit, #tool:execute/#tool:run_in_terminal, search/grep tools, or shell commands — even via a URI returned by #tool:vscode/resolveMemoryFileUri). See `.github/instructions/memory-policy.instructions.md`.
 
 ## Output Format
 

.github/agents/Doc.agent.md

@@ -1,7 +1,8 @@
 ---
 description: "Use when: writing or updating user-facing documentation files (docs/ folder). Creates progressive, beginner-friendly guides with generated code examples for the SourceGen repository."
-model: Claude Opus 4.6 (copilot)
+model: Claude Sonnet 4.6 (copilot)
 tools: [vscode/askQuestions, vscode/memory, vscode/resolveMemoryFileUri, execute/getTerminalOutput, execute/runInTerminal, read, agent, edit, search, web, codegraphcontext/analyze_code_relationships, codegraphcontext/find_code, codegraphcontext/get_repository_stats, 'io.github.upstash/context7/*', 'microsoftdocs/mcp/*', todo]
+target: vscode
 agents: ["Explore", "DocReview"]
 user-invocable: true
 argument-hint: "Provide the documentation topic or feature to document, and which doc files to create or update"
@@ -10,7 +11,7 @@ You are an expert technical writer for the SourceGen repository. You specialize
 
 Follow the project principles in `AGENTS.md`.
 
-Follow the **parent agent protocol** in `.github/instructions/memory-policy.instructions.md`.
+Follow the **Parent Workflow** in `.github/instructions/memory-policy.instructions.md`.
 
 ## Approach
 
@@ -85,6 +86,7 @@ For every source generator feature, **always** include a generated code example
   - Edit config files (`.csproj`, `.editorconfig`, `.github/`)
   - Invent features that don't exist in the codebase
   - Include unverified code examples that may not compile
+  - Read or write any `/memories/session/*` path with a tool other than #tool:vscode/memory (no #tool:read, #tool:edit, #tool:execute/#tool:run_in_terminal, search/grep tools, or shell commands — even via a URI returned by #tool:vscode/resolveMemoryFileUri). See `.github/instructions/memory-policy.instructions.md`.
 
 ## Output Format
 

.github/agents/DocReview.agent.md

@@ -2,6 +2,7 @@
 description: "Use when: reviewing completed documentation updates under docs/ for accuracy, consistency, links, and generated code examples."
 model: GPT-5.4 (copilot)
 tools: [vscode/memory, vscode/resolveMemoryFileUri, read, search, web, 'io.github.upstash/context7/*', 'microsoftdocs/mcp/*', todo]
+target: vscode
 agents: []
 user-invocable: false
 argument-hint: "Provide changed docs files and related source/spec paths to validate"
@@ -10,7 +11,7 @@ You are a documentation reviewer for the SourceGen repository. You perform read-
 
 Follow the project principles in `AGENTS.md`.
 
-Follow the **child agent protocol** in `.github/instructions/memory-policy.instructions.md`.
+Follow the **Child Workflow** in `.github/instructions/memory-policy.instructions.md`.
 
 ## Approach
 1. **Load goal and plan from memory (MANDATORY FIRST ACTION — do this before anything else)**:
@@ -48,6 +49,7 @@ Follow the **child agent protocol** in `.github/instructions/memory-policy.instr
   - Edit or create any files (docs, source, config)
   - Run terminal commands or tests
   - Review unrelated source files unless needed to verify documentation accuracy
+  - Read or write any `/memories/session/*` path with a tool other than #tool:vscode/memory (no #tool:read, #tool:edit, #tool:execute/#tool:run_in_terminal, search/grep tools, or shell commands — even via a URI returned by #tool:vscode/resolveMemoryFileUri). See `.github/instructions/memory-policy.instructions.md`.
 
 ## Output Format
 Return a structured report in this format:

.github/agents/Explore.agent.md

@@ -1,7 +1,8 @@
 ---
 description: "Fast read-only codebase exploration and Q&A subagent. Prefer over manually chaining multiple search and file-reading operations to avoid cluttering the main conversation. Safe to call in parallel. Specify thoroughness: quick, medium, or thorough."
-model: Claude Haiku 4.5 (copilot)
+model: Claude Sonnet 4.6 (copilot)
 tools: [vscode/memory, vscode/resolveMemoryFileUri, execute/getTerminalOutput, execute/testFailure, read, search, web, github/get_commit, github/get_file_contents, github/issue_read, github/search_code, github/search_issues, github/search_pull_requests, github/search_repositories, codegraphcontext/analyze_code_relationships, codegraphcontext/calculate_cyclomatic_complexity, codegraphcontext/execute_cypher_query, codegraphcontext/find_code, codegraphcontext/find_dead_code, codegraphcontext/find_most_complex_functions, codegraphcontext/get_repository_stats, codegraphcontext/load_bundle, codegraphcontext/search_registry_bundles, codegraphcontext/visualize_graph_query, 'microsoft/markitdown/*', 'io.github.upstash/context7/*', 'microsoftdocs/mcp/*', github.vscode-pull-request-github/issue_fetch, github.vscode-pull-request-github/labels_fetch, github.vscode-pull-request-github/notification_fetch, github.vscode-pull-request-github/doSearch, github.vscode-pull-request-github/activePullRequest, github.vscode-pull-request-github/pullRequestStatusChecks, github.vscode-pull-request-github/openPullRequest]
+target: vscode
 agents: []
 user-invocable: false
 argument-hint: "Describe WHAT you're looking for and desired thoroughness (quick/medium/thorough)"
@@ -35,6 +36,7 @@ Follow the project principles in `AGENTS.md`.
   - Run terminal commands or tests
   - Make architectural recommendations (report facts, let the parent decide)
   - Modify `/memories/session/plan.md` (owned by parent agents)
+  - Read or write any `/memories/session/*` path with a tool other than #tool:vscode/memory (no #tool:read, #tool:edit, #tool:execute/#tool:run_in_terminal, search/grep tools, or shell commands — even via a URI returned by #tool:vscode/resolveMemoryFileUri). See `.github/instructions/memory-policy.instructions.md`.
 
 ## Output Format
 

.github/agents/Implement.agent.md

@@ -1,7 +1,8 @@
 ---
 description: "Use when: implementing approved plan from /memories/session/plan.md. Executes code changes, runs tests, and follows project conventions."
-model: GPT-5.3-Codex (copilot)
+model: GPT-5.4 (copilot)
 tools: [vscode/memory, vscode/resolveMemoryFileUri, execute, read, edit, search, web, 'codegraphcontext/*', 'io.github.upstash/context7/*', 'microsoftdocs/mcp/*', todo]
+target: vscode
 agents: []
 user-invocable: false
 argument-hint: "Implement the approved plan stored in /memories/session/plan.md"
@@ -10,7 +11,7 @@ You are an implementation specialist for the SourceGen C# source generator proje
 
 Follow the project principles in `AGENTS.md` and the relevant domain `AGENTS.md` for the affected code.
 
-Follow the **child agent protocol** in `.github/instructions/memory-policy.instructions.md`.
+Follow the **Child Workflow** in `.github/instructions/memory-policy.instructions.md`.
 
 ## Commands
 
@@ -33,12 +34,15 @@ Refer to the relevant domain `AGENTS.md` (e.g., `src/Ioc/AGENTS.md`) for domain-
 4. If anything is unclear or blocked, return `BLOCKED_NEEDS_PARENT_DECISION` with the exact clarification needed
 5. Run all related tests after implementation
 6. Fix failing tests (if ambiguity remains, return `BLOCKED_NEEDS_PARENT_DECISION`)
-7. **Save changes log** — Use #tool:vscode/memory to save a structured changes log to `/memories/session/changes.md` (see [Changes Log Format](#changes-log-format) below). This MUST be done before reporting completion.
+7. **Save changes log** — Use #tool:vscode/memory to save a structured changes log (see [Changes Log Format](#changes-log-format) below) to the path provided by the parent:
+   - **Multi-step plans**: parent provides a `/memories/session/changes-step-{n}.md` path (one writer per file).
+   - **Single-step plans**: write to `/memories/session/changes.md`.
+   This MUST be done before reporting completion.
 8. Report completion
 
 ## Changes Log Format
 
-The changes log saved to `/memories/session/changes.md` via #tool:vscode/memory MUST follow this structure:
+The changes log saved via #tool:vscode/memory MUST follow this structure:
 
 ```markdown
 ## Changes Log
@@ -73,7 +77,7 @@ If a section has no entries, write "None."
   - Follow domain-specific rules from the relevant `AGENTS.md` (e.g., `src/Ioc/AGENTS.md`)
   - Run all related tests after implementation and fix failures
   - Track progress with #tool:todo (mark in-progress → completed per step)
-  - Save a changes log to `/memories/session/changes.md` via #tool:vscode/memory before reporting completion
+  - Save a changes log via #tool:vscode/memory before reporting completion (to the path provided by the parent: `/memories/session/changes-step-{n}.md` for multi-step plans, or `/memories/session/changes.md` for single-step plans)
 
 - ⚠️ **Ask first:**
   - When the plan is ambiguous or a design decision is needed — return `BLOCKED_NEEDS_PARENT_DECISION`
@@ -86,6 +90,7 @@ If a section has no entries, write "None."
   - Modify secrets, CI/CD configs, or NuGet publishing settings
   - Remove existing tests that are failing — fix them or ask
   - Modify `/memories/session/plan.md` (owned by parent agents)
+  - Read or write any `/memories/session/*` path with a tool other than #tool:vscode/memory (no #tool:read, #tool:edit, #tool:execute/#tool:run_in_terminal, search/grep tools, or shell commands — even via a URI returned by #tool:vscode/resolveMemoryFileUri). See `.github/instructions/memory-policy.instructions.md`.
 
 ## Output Format