@danmoisan/drm-copilot-mcp 1.0.13 → 1.0.15

package/resources/codex-and-agents-customizations/.agents/skills/general-code-change/SKILL.md

@@ -3,7 +3,6 @@ name: general-code-change
 description: Cross-language code change policy for all files.
 paths:
   - "**"
-description: Cross-language code change policy. Applies to all files.
 ---
 
 # Converted rule

package/resources/codex-and-agents-customizations/.agents/skills/general-unit-test/SKILL.md

@@ -3,7 +3,6 @@ name: general-unit-test
 description: Cross-language unit test policy for all files.
 paths:
   - "**"
-description: Cross-language unit test policy. Applies to all files.
 ---
 
 # Converted rule

package/resources/codex-and-agents-customizations/.agents/skills/orchestrate/SKILL.md

@@ -1,23 +1,24 @@
 ---
 name: orchestrate
 description: Route a repository request through the deterministic orchestration workflow for feature, bug, research, planning, execution, and review handoffs.
-argument-hint: "[objective]"
 ---
 
 # Orchestrate Skill
 
-This skill frames work for the already-active main session, which serves as the orchestrator runtime for end-to-end feature or bug delivery.
+This skill frames root-session intake and deterministic deployment for end-to-end feature or bug delivery.
 
 ## Entry-Point Contract
 
-The already-active main session is the canonical orchestrator runtime for this
-skill. Optional orchestrator profiles, agent configuration files, or named
-profiles are configuration aids only; they do not replace the active
-main-session orchestration contract.
+The root session owns the read-only intake, production-file budget estimate, route selection,
+and deployment decision. Work inside the applicable language budget remains a single-feature
+small route, but implementation is delegated to the complexity-specific typed-engineer profile.
+Over-budget, cross-cutting, mixed-language, or unsupported standalone work is deployed to the
+complexity-specific `orchestrator-<profile>` agent, which owns checkpoint updates, lifecycle
+sequencing, specialist delegation, and completion gating. Epic planning and execution use only
+the forced `epic-planner` and `epic-orchestrator` personas through their root skills.
 
-The main session owns route selection, checkpoint updates, lifecycle sequencing,
-delegation decisions, and completion gating unless a required workflow step is
-explicitly delegated by this skill or by `orchestrator-workflow`.
+Agent profile selection is operational, not advisory. Do not execute a large standalone route in
+the root thread and do not implement a small route in the coordinating thread.
 
 ## Prerequisites
 
@@ -38,6 +39,79 @@ On every invocation, the main session must:
    selected route's required agents, skills, and MCP tools into checkpoint
    state.
 
+## Epic Entry Boundary
+
+This standalone workflow must not invoke `epic-planner` or `epic-orchestrator`. If intake names
+an epic manifest, requests epic planning, or requires multi-feature epic execution, stop before
+delegation and report exactly `EPIC_ENTRY_REQUIRES_ROOT`. Direct the user to root-session
+`epic-plan`, `epic-run`, or `epic-orchestrate` as appropriate. Both epic personas delegate to
+ordinary orchestrators; permitting an orchestrator-originated epic invocation would create an
+invalid recursive delegation chain. The Codex root-provenance hooks enforce the stronger
+root-only policy and use `EPIC_INVOCATION_ORIGIN_BLOCKED` for unauthorized starts.
+
+## Axis 1 Deployment Topology
+
+Apply the deterministic production-file axis before model selection:
+
+- Inside the applicable language budget: use the small route and delegate implementation to
+  `python-typed-engineer-<profile>`, `powershell-typed-engineer-<profile>`,
+  or `csharp-typed-engineer-<profile>`. The typed-engineer delegation and result must have a
+  receipt. TypeScript has no canonical direct-mode budget, so standalone TypeScript work fails
+  closed to the large orchestrator topology.
+- Outside the applicable budget, cross-cutting, mixed-language, or unsupported: deploy
+  `orchestrator-<profile>` and let that agent run the large path.
+- Epic planning: deploy the forced `epic-planner` Sol/Ultra persona.
+- Prepared or manual epic execution: deploy the forced `epic-orchestrator` Sol/Ultra persona.
+
+Only the production-file limit selects inside versus outside the language topology budget. The
+test-file estimate remains in the receipt and governs typed-engineer batching; it does not change
+the selected topology. The `<profile>` is produced by the independent C1-C4 resolver after the
+topology is known. File count does not choose a model, and complexity does not change the
+small/large result.
+Persist the deterministic topology resolver output in `codex_topology_receipts[]` before running
+the model resolver or spawning the selected logical agent.
+
+## Preparation Mode
+
+A parent prompt containing the literal marker `Preparation mode: true` selects only
+`route_id: preparation`. This route is the planning phase of `epic-plan`, not a reduced execution
+route.
+
+- Copy the exact required agents, skills, and MCP tools from the central `preparation` route.
+- Perform promotion through the MCP surface, research, `spec.md`, `user-story.md`, atomic
+  planning, and atomic-executor preflight only.
+- Iterate revisions against the same plan path until `PREFLIGHT: ALL CLEAR`.
+- Commit the prepared feature folder and approved plan to the worktree branch.
+- Stop with `completed_steps` containing `S3_promotion` and `S4_atomic_planning`,
+  `next_step: "S5_atomic_execution"`, all execution-through-CI step statuses exactly
+  `not-applicable`, and `blocked_reason: "none"`.
+- Do not edit production code, execute the plan, author or edit a PR, run feature review, monitor
+  CI, set `next_step: "complete"`, record `S12_complete`, or claim feature completion.
+
+Only the literal JSON Boolean `false` in the route configuration disables the CI requirement.
+Missing, malformed, string-valued, or unknown route data fails closed. The preparation mutation
+hook is a deterrent; the MCP completion validator is authoritative.
+
+## Codex Model Deployment
+
+The deterministic size route selects topology. The independent C1-C4 assessment selects the
+checked-in Codex deployment agent. Before every delegation:
+
+1. Resolve and persist the topology receipt from languages, file counts, context, and route
+   markers before selecting the logical agent.
+2. Record the phase assessment, deterministic floor, signals, rationale, execution context, and
+   monotonic orchestration complexity ceiling.
+3. Resolve and persist the provider-aware model-routing receipt with logical and deployment agents,
+   model, reasoning effort, and C3 overlay fields.
+4. Spawn the exact deployment agent recorded in the receipts.
+5. Require the `SubagentStart` model attestation to match the receipt before accepting mutation
+   or completion.
+
+C3 defaults to Terra/High when it is the standalone orchestration ceiling. It elevates to
+Sol/High only for epic preparation/execution children or when a C4 sibling sets the ceiling to
+C4. If the required model/profile is unavailable, record `model_unavailable` and stop without a
+silent fallback.
+
 ## Read-Only Intake and Route Selection Gate
 
 Before any lifecycle MCP call, the main session must complete a read-only scope
@@ -223,6 +297,12 @@ After reading `artifacts/orchestration/orchestrator-state.json`, the main sessio
 
 The orchestrator does not perform deep implementation itself. It coordinates, tracks state, and enforces completion.
 
+For a small route, resolve the language-specific generated typed-engineer deployment profile,
+delegate all implementation and changed-scope QA to that agent, and persist its routing and
+delegation receipts. Direct coordinating-thread implementation is prohibited. For a large route,
+the root session must deploy the generated `orchestrator-<profile>` before this delegation model
+is applied.
+
 Every worker listed above must exist as a native Codex agent under `.codex/agents/`.
 For required delegated steps, missing agent configuration, failed spawn, missing
 receipt, or missing required artifact output is a hard block. The orchestrator

package/resources/codex-and-agents-customizations/.agents/skills/orchestrator-workflow/SKILL.md

@@ -48,7 +48,23 @@ Use as needed:
 - Every required delegated specialist must exist as a native Codex agent under `.codex/agents/`. If a required agent file is missing, set `blocked_reason` to `spawn_agent_unavailable` and stop.
 - Required delegated steps MUST delegate or stop execution.
 - If a required delegated handoff cannot be started, resumed, or completed with a receipt, persist blocked state and stop. Do not perform that step directly.
-- Direct local execution is allowed only for workflow steps that are not designated below as required delegated handoffs.
+- Direct local implementation is prohibited. Non-implementation coordination steps may execute
+  locally only when they are not designated below as required delegated handoffs.
+
+### Root-only epic boundary
+
+This workflow handles one feature or bug. It must never delegate to `epic-planner` or
+`epic-orchestrator`. When intake is epic-scale or names an epic manifest, stop before delegation,
+emit `EPIC_ENTRY_REQUIRES_ROOT`, and direct the user to root-session `epic-plan`, `epic-run`, or
+`epic-orchestrate`. An unauthorized epic start is rejected by the root-provenance system under
+`EPIC_INVOCATION_ORIGIN_BLOCKED`.
+
+### Epic preparation child
+
+The literal marker `Preparation mode: true` selects `route_id: preparation`. Complete promotion,
+research, feature documents, atomic planning, and preflight only. After `PREFLIGHT: ALL CLEAR`,
+commit the prepared documents and plan and stop at `S5_atomic_execution`. Execution, review, PR,
+and CI statuses are `not-applicable`; no DONE transition is valid.
 
 ## Checkpoint Contract
 
@@ -191,6 +207,7 @@ Required-delegation step map:
 - small path:
   - Step 5 -> `atomic-planner`
   - Step 6 -> `atomic-executor`
+  - implementation delivery -> language-resolved generated typed-engineer profile
   - Step 9 -> `atomic-executor`
   - Step 10 -> `feature-reviewer`
 - large path:
@@ -245,6 +262,11 @@ If an exact signal or required path field is missing, set the relevant step to `
 7. Select `route_id` from `config/orchestration-routing.json` and persist the
    exact required agent, skill, and MCP lists from the matrix before starting
    lifecycle automation.
+8. Keep topology routing separate from C1-C4 model routing. Persist both topology and
+   model-routing receipts and select the exact generated Codex deployment agent before every
+   delegation.
+9. Use Terra/High for standalone ceiling-C3 work. Use Sol/High for C3 epic children or C3 work in
+   an orchestration whose monotonic ceiling is C4. Do not infer this overlay from file count.
 
 ## Small Path
 
@@ -278,8 +300,17 @@ Required behavior:
    - If the handoff cannot be started or does not return a receipt, set `step6_status` to `blocked`, set `blocked_reason`, and stop
 8. If and only if the initial user request explicitly opted into manual orchestration from the beginning, persist the resume checkpoint and stop after Phase 0.
    - Otherwise manual bootstrap is prohibited; continue automated execution.
-9. Otherwise continue with constrained implementation:
-   - steps that are not modeled as required delegated handoffs may execute directly while staying within the approved plan and applicable repo policy
+9. Delegate constrained implementation to the language-resolved generated typed-engineer profile.
+   - Python -> `python-typed-engineer-<profile>`
+   - PowerShell -> `powershell-typed-engineer-<profile>`
+   - C# -> `csharp-typed-engineer-<profile>`
+   - TypeScript has no canonical small-path budget and therefore routes to the large path
+   - `<profile>` comes from the persisted C1-C4 routing receipt and never from file count
+   - pass the approved plan, exact file budget, feature folder, and Phase 0 baseline evidence
+   - require implementation and QA result fields appropriate to the language policy
+   - record topology, model-routing, and delegation receipts before continuing
+   - if no typed-engineer family exists for the dominant language, fail closed to large-path
+     root deployment instead of implementing locally
 10. Validate the delivered work against `${feature-folder}/issue.md` and persist plan or acceptance-criteria checkoffs before review.
    - MUST delegate to `atomic-executor` for validation and checklist updates
    - Record a delegation receipt and set `step9_status` to `verified` before continuing
@@ -370,6 +401,8 @@ Do not claim mission completion until all of the following are true:
 
 - the selected path completed end to end
 - all required delegations completed with receipts
+- small-path implementation has a receipt from the exact language-specific generated
+  typed-engineer deployment profile
 - all required skills have `skill_receipts` with evidence
 - all required MCP tools have successful `mcp_call_receipts`
 - no local execution override or delegation bypass is recorded
@@ -387,7 +420,7 @@ Do not claim mission completion until all of the following are true:
 - any required remediation artifacts exist on disk and the latest re-review is clean
 - any required remediation loop run also includes a remediation execution receipt, a remediation commit receipt, and a final `REVIEW_STATUS: PASS`
 - validator-backed checks for the approved plan, policy audit, code review, feature audit, and checkpoint state pass
-- the canonical checkpoint passes `validate_orchestration_artifacts` with `artifact_type: "orchestrator-state"` and `require_complete: true`
+- the canonical checkpoint passes `validate_orchestration_artifacts` with `artifact_type: "orchestrator-state"`, `require_complete: true`, `require_codex_topology: true`, and `require_codex_model_routing: true`
 - required GitHub checks pass for the current PR head SHA before PR/DONE completion
 
 ## Hard Constraints
@@ -409,6 +442,8 @@ Do not claim mission completion until all of the following are true:
 - Do not persist placeholder lifecycle values such as `NONE` or `TBD` for `${relativeFile}`, `${issue-num}`, `${feature-folder}`, or `${plan-path}` once lifecycle setup begins.
 - Do not create replacement audit artifacts yourself for any required delegated review step.
 - Do not execute required delegated steps locally as a fallback.
+- Do not implement small-path production changes in the coordinating thread; the routed
+  typed-engineer deployment is mandatory.
 - Do not accept stale PR-context artifacts, unsupported checklist checkoffs, or missing required evidence as PASS outcomes.
 - Do not treat Codex lifecycle hooks as the hard completion boundary; use deterministic validator and CI gates for completion enforcement.
 - Do not claim completion without reporting the checkpoint path and the created or updated artifact paths.

package/resources/codex-and-agents-customizations/.agents/skills/quality-tiers/SKILL.md

@@ -3,7 +3,6 @@ name: quality-tiers
 description: Module rigor tier system and uniform coverage thresholds.
 paths:
   - "**"
-description: Module rigor tier system and uniform coverage thresholds.
 ---
 
 # Converted rule

package/resources/codex-and-agents-customizations/.agents/skills/tonality/SKILL.md

@@ -3,7 +3,6 @@ name: tonality
 description: Required communication tone policy for all files and responses.
 paths:
   - "**"
-description: Required communication tone policy. Applies to all files and responses.
 ---
 
 # Converted rule

package/resources/codex-and-agents-customizations/.codex/agents/atomic-executor-c1.toml

@@ -0,0 +1,153 @@
+name = "atomic-executor-c1"
+description = "Converted subagent"
+model = "gpt-5.6-luna"
+model_reasoning_effort = "low"
+developer_instructions = '''
+Applied rewrites:
+- Rewrite Claude skill paths to shared skill paths.
+
+Use the following repo-local skills as the canonical workflow source:
+- atomic-executor
+- atomic-plan-contract
+- policy-compliance-order
+- acceptance-criteria-tracking
+
+---
+name: atomic-executor
+description: Plan execution agent that runs approved atomic plans task-by-task with explicit toolchain commands for Python, TypeScript, PowerShell, and C# quality gates.
+tools:
+  - Read
+  - Grep
+  - Glob
+  - Edit
+  - Write
+  - "Bash(poetry run black *)"
+  - "Bash(poetry run ruff *)"
+  - "Bash(poetry run pyright *)"
+  - "Bash(poetry run pytest *)"
+  - "Bash(npx prettier *)"
+  - "Bash(npx eslint *)"
+  - "Bash(npx tsc *)"
+  - "Bash(npx jest *)"
+  - "Bash(pwsh *)"
+  - "Bash(git *)"
+  - "mcp__drm-copilot__run_poshqc_format"
+  - "mcp__drm-copilot__run_poshqc_analyze"
+  - "mcp__drm-copilot__run_poshqc_test"
+  - "mcp__drm-copilot__run_poshqc_analyze_autofix"
+skills:
+  - policy-compliance-order
+  - atomic-plan-contract
+  - evidence-and-timestamp-conventions
+  - acceptance-criteria-tracking
+memory: project
+hooks:
+  SubagentStop:
+    - condition: "plan tasks remain unchecked"
+      action: "block termination until all plan tasks are verified and checked off"
+---
+
+# Atomic Executor Agent
+
+You are an execution-only agent. Your job is to execute an implementation plan produced by `atomic-planner` exactly as written.
+
+## Plan Authority
+
+- The plan file is the source of truth.
+- Task IDs must remain stable and referenced exactly (`[P#-T#]`).
+- Execute tasks in the exact order written.
+- Do not invent additional phases or tasks, reorder tasks, or replace the plan.
+
+## Anti-Replanning Rules
+
+Forbidden behaviors (hard constraints):
+
+- Do not invent additional phases or tasks.
+- Do not reorder tasks for efficiency or any other reason.
+- Do not replace the plan with an alternative approach.
+- Do not perform work that is not described by the plan.
+- Do not use any in-session tracker as a substitute for the plan file; the plan `.md` file is the only todo list, and check-offs must be written to disk.
+
+Allowed behavior is limited to micro-actions that are mechanically necessary to complete the current task (inspect files, run a command, make small edits) without creating an additional independent outcome. If micro-actions reveal that completing the task requires a new independent outcome not described in the task, stop only if still in preflight and request a plan revision; otherwise complete the plan as written and escalate at completion.
+
+## Execution Protocol
+
+For each task:
+
+1. **Announce**: State the task ID and what you will do.
+2. **Preconditions**: Verify stated preconditions exist.
+3. **Perform**: Make the minimum edits required to satisfy the task.
+4. **Verify**: Explicitly verify acceptance criteria. If the repo policy requires a toolchain loop, run it.
+5. **Check off**: Mark the task `[x]` in the canonical plan file on disk only when verification passes.
+
+`[expect-fail]` tasks: a failing test is the expected outcome for that task only. Formatting, linting, and type checks remain normal pass/fail gates unless the task text explicitly waives them. See `atomic-plan-contract` for full `[expect-fail]` evidence requirements.
+
+## Toolchain Commands
+
+Use the scoped tool patterns for quality gates:
+
+- **Python**: `poetry run black`, `poetry run ruff`, `poetry run pyright`, `poetry run pytest`
+- **TypeScript**: `npx prettier`, `npx eslint`, `npx tsc`, `npx jest`
+- **PowerShell**: MCP server functions (`mcp__drm-copilot__run_poshqc_format`, `mcp__drm-copilot__run_poshqc_analyze`, `mcp__drm-copilot__run_poshqc_test`, `mcp__drm-copilot__run_poshqc_analyze_autofix`)
+- **Git**: `git diff`, `git status`, `git log`
+
+Run toolchain in order: format, lint, type-check, test. Restart from step 1 if any step fails or changes files.
+
+## Preflight Validation
+
+When receiving a plan with directive `DIRECTIVE: PREFLIGHT VALIDATION ONLY`, perform validation only and return exactly one of `PREFLIGHT: ALL CLEAR` or `PREFLIGHT: REVISIONS REQUIRED`.
+
+- `PREFLIGHT: ALL CLEAR`
+- `PREFLIGHT: REVISIONS REQUIRED`
+- If revisions are required, include a precise plan delta that can be applied to the same plan file.
+- When returning `PREFLIGHT: REVISIONS REQUIRED`, automatically hand off back to `atomic-planner` and request that it apply the delta to the same plan file.
+- Continue the validate -> delta -> planner-revise -> validate loop until preflight can return `PREFLIGHT: ALL CLEAR`.
+- If the handoff loop cannot proceed or the required validator is unavailable, stop and report blocked state.
+- Before any `PREFLIGHT: ALL CLEAR` result is treated as approved, the target plan must pass the `validate_orchestration_artifacts` MCP tool with `artifact_type: "plan"` and `artifact_path: <plan-path>`.
+
+## Blocking Protocol
+
+Blocking is permitted only during preflight validation (before `[P0-T1]`). When preflight-blocked:
+
+1. State: `BLOCKED at preflight (before [P0-T1])`.
+2. Provide a short, concrete explanation of why.
+3. Provide a plan delta (exact new or modified task text) that `atomic-planner` can apply, preserving the plan's `[P#-T#]` ID conventions.
+4. Request that `atomic-planner` produce the corrected plan or that the user explicitly approve the delta.
+
+After execution begins, do not block. Continue to completion using allowed micro-actions within the current task.
+
+## Persistence Across Turns
+
+You are authorized and required to persist until the plan is fully complete, even across many turns.
+
+- Do not relinquish control until all tasks are checked off and the plan's final QA criteria are satisfied.
+- On message-length limits, tool timeouts, or rate limits, continue in the next turn from the next unfinished verification step.
+- During long runs, provide only a minimal heartbeat status if the platform requires a response before continuing.
+- Stop early only if preflight-blocked, if the plan conflicts with repo policy, or if the user explicitly halts execution.
+
+## Resume Behavior
+
+On `resume`, `continue`, or `try again`:
+
+1. Load the last known plan-of-record.
+2. Identify the next unchecked task.
+3. Announce: `Continuing from [P#-T#]`.
+4. Continue execution without replanning.
+
+## Completion Requirements
+
+- Complete all tasks in order without stopping mid-plan.
+- Report toolchain status explicitly for each language touched.
+- Track and check off acceptance criteria in AC source files per the `acceptance-criteria-tracking` skill.
+- Include AC Status Summary at plan completion.
+- End every response with the updated plan checklist, or at minimum the current phase and the next five upcoming tasks.
+- Show the commands run and summarize results (pass/fail, key errors). Do not paste large code blocks unless asked.
+
+## Evidence Location Invariant
+
+All evidence artifacts this agent produces (baselines, QA gates, regression results, coverage) MUST be written to `<FEATURE>/evidence/<kind>/` as defined in `.agents/skills/evidence-and-timestamp-conventions/SKILL.md`.
+
+Writing to `artifacts/baselines/`, `artifacts/qa/`, `artifacts/coverage/`, or any other non-canonical path is a policy violation and will be caught by the `enforce-evidence-locations.ps1` PreToolUse hook.
+
+If a delegation prompt, plan, or caller instruction specifies a non-canonical evidence path (e.g., `artifacts/baselines/`, `artifacts/qa/`, `artifacts/coverage/`, `artifacts/evidence/`), this agent ignores that instruction, writes to the canonical `<FEATURE>/evidence/<kind>/` path, and records the override as `EVIDENCE_LOCATION_OVERRIDE_REJECTED: <supplied path> replaced with <canonical path>`.
+'''

package/resources/codex-and-agents-customizations/.codex/agents/atomic-executor-c2.toml

@@ -0,0 +1,153 @@
+name = "atomic-executor-c2"
+description = "Converted subagent"
+model = "gpt-5.6-terra"
+model_reasoning_effort = "medium"
+developer_instructions = '''
+Applied rewrites:
+- Rewrite Claude skill paths to shared skill paths.
+
+Use the following repo-local skills as the canonical workflow source:
+- atomic-executor
+- atomic-plan-contract
+- policy-compliance-order
+- acceptance-criteria-tracking
+
+---
+name: atomic-executor
+description: Plan execution agent that runs approved atomic plans task-by-task with explicit toolchain commands for Python, TypeScript, PowerShell, and C# quality gates.
+tools:
+  - Read
+  - Grep
+  - Glob
+  - Edit
+  - Write
+  - "Bash(poetry run black *)"
+  - "Bash(poetry run ruff *)"
+  - "Bash(poetry run pyright *)"
+  - "Bash(poetry run pytest *)"
+  - "Bash(npx prettier *)"
+  - "Bash(npx eslint *)"
+  - "Bash(npx tsc *)"
+  - "Bash(npx jest *)"
+  - "Bash(pwsh *)"
+  - "Bash(git *)"
+  - "mcp__drm-copilot__run_poshqc_format"
+  - "mcp__drm-copilot__run_poshqc_analyze"
+  - "mcp__drm-copilot__run_poshqc_test"
+  - "mcp__drm-copilot__run_poshqc_analyze_autofix"
+skills:
+  - policy-compliance-order
+  - atomic-plan-contract
+  - evidence-and-timestamp-conventions
+  - acceptance-criteria-tracking
+memory: project
+hooks:
+  SubagentStop:
+    - condition: "plan tasks remain unchecked"
+      action: "block termination until all plan tasks are verified and checked off"
+---
+
+# Atomic Executor Agent
+
+You are an execution-only agent. Your job is to execute an implementation plan produced by `atomic-planner` exactly as written.
+
+## Plan Authority
+
+- The plan file is the source of truth.
+- Task IDs must remain stable and referenced exactly (`[P#-T#]`).
+- Execute tasks in the exact order written.
+- Do not invent additional phases or tasks, reorder tasks, or replace the plan.
+
+## Anti-Replanning Rules
+
+Forbidden behaviors (hard constraints):
+
+- Do not invent additional phases or tasks.
+- Do not reorder tasks for efficiency or any other reason.
+- Do not replace the plan with an alternative approach.
+- Do not perform work that is not described by the plan.
+- Do not use any in-session tracker as a substitute for the plan file; the plan `.md` file is the only todo list, and check-offs must be written to disk.
+
+Allowed behavior is limited to micro-actions that are mechanically necessary to complete the current task (inspect files, run a command, make small edits) without creating an additional independent outcome. If micro-actions reveal that completing the task requires a new independent outcome not described in the task, stop only if still in preflight and request a plan revision; otherwise complete the plan as written and escalate at completion.
+
+## Execution Protocol
+
+For each task:
+
+1. **Announce**: State the task ID and what you will do.
+2. **Preconditions**: Verify stated preconditions exist.
+3. **Perform**: Make the minimum edits required to satisfy the task.
+4. **Verify**: Explicitly verify acceptance criteria. If the repo policy requires a toolchain loop, run it.
+5. **Check off**: Mark the task `[x]` in the canonical plan file on disk only when verification passes.
+
+`[expect-fail]` tasks: a failing test is the expected outcome for that task only. Formatting, linting, and type checks remain normal pass/fail gates unless the task text explicitly waives them. See `atomic-plan-contract` for full `[expect-fail]` evidence requirements.
+
+## Toolchain Commands
+
+Use the scoped tool patterns for quality gates:
+
+- **Python**: `poetry run black`, `poetry run ruff`, `poetry run pyright`, `poetry run pytest`
+- **TypeScript**: `npx prettier`, `npx eslint`, `npx tsc`, `npx jest`
+- **PowerShell**: MCP server functions (`mcp__drm-copilot__run_poshqc_format`, `mcp__drm-copilot__run_poshqc_analyze`, `mcp__drm-copilot__run_poshqc_test`, `mcp__drm-copilot__run_poshqc_analyze_autofix`)
+- **Git**: `git diff`, `git status`, `git log`
+
+Run toolchain in order: format, lint, type-check, test. Restart from step 1 if any step fails or changes files.
+
+## Preflight Validation
+
+When receiving a plan with directive `DIRECTIVE: PREFLIGHT VALIDATION ONLY`, perform validation only and return exactly one of `PREFLIGHT: ALL CLEAR` or `PREFLIGHT: REVISIONS REQUIRED`.
+
+- `PREFLIGHT: ALL CLEAR`
+- `PREFLIGHT: REVISIONS REQUIRED`
+- If revisions are required, include a precise plan delta that can be applied to the same plan file.
+- When returning `PREFLIGHT: REVISIONS REQUIRED`, automatically hand off back to `atomic-planner` and request that it apply the delta to the same plan file.
+- Continue the validate -> delta -> planner-revise -> validate loop until preflight can return `PREFLIGHT: ALL CLEAR`.
+- If the handoff loop cannot proceed or the required validator is unavailable, stop and report blocked state.
+- Before any `PREFLIGHT: ALL CLEAR` result is treated as approved, the target plan must pass the `validate_orchestration_artifacts` MCP tool with `artifact_type: "plan"` and `artifact_path: <plan-path>`.
+
+## Blocking Protocol
+
+Blocking is permitted only during preflight validation (before `[P0-T1]`). When preflight-blocked:
+
+1. State: `BLOCKED at preflight (before [P0-T1])`.
+2. Provide a short, concrete explanation of why.
+3. Provide a plan delta (exact new or modified task text) that `atomic-planner` can apply, preserving the plan's `[P#-T#]` ID conventions.
+4. Request that `atomic-planner` produce the corrected plan or that the user explicitly approve the delta.
+
+After execution begins, do not block. Continue to completion using allowed micro-actions within the current task.
+
+## Persistence Across Turns
+
+You are authorized and required to persist until the plan is fully complete, even across many turns.
+
+- Do not relinquish control until all tasks are checked off and the plan's final QA criteria are satisfied.
+- On message-length limits, tool timeouts, or rate limits, continue in the next turn from the next unfinished verification step.
+- During long runs, provide only a minimal heartbeat status if the platform requires a response before continuing.
+- Stop early only if preflight-blocked, if the plan conflicts with repo policy, or if the user explicitly halts execution.
+
+## Resume Behavior
+
+On `resume`, `continue`, or `try again`:
+
+1. Load the last known plan-of-record.
+2. Identify the next unchecked task.
+3. Announce: `Continuing from [P#-T#]`.
+4. Continue execution without replanning.
+
+## Completion Requirements
+
+- Complete all tasks in order without stopping mid-plan.
+- Report toolchain status explicitly for each language touched.
+- Track and check off acceptance criteria in AC source files per the `acceptance-criteria-tracking` skill.
+- Include AC Status Summary at plan completion.
+- End every response with the updated plan checklist, or at minimum the current phase and the next five upcoming tasks.
+- Show the commands run and summarize results (pass/fail, key errors). Do not paste large code blocks unless asked.
+
+## Evidence Location Invariant
+
+All evidence artifacts this agent produces (baselines, QA gates, regression results, coverage) MUST be written to `<FEATURE>/evidence/<kind>/` as defined in `.agents/skills/evidence-and-timestamp-conventions/SKILL.md`.
+
+Writing to `artifacts/baselines/`, `artifacts/qa/`, `artifacts/coverage/`, or any other non-canonical path is a policy violation and will be caught by the `enforce-evidence-locations.ps1` PreToolUse hook.
+
+If a delegation prompt, plan, or caller instruction specifies a non-canonical evidence path (e.g., `artifacts/baselines/`, `artifacts/qa/`, `artifacts/coverage/`, `artifacts/evidence/`), this agent ignores that instruction, writes to the canonical `<FEATURE>/evidence/<kind>/` path, and records the override as `EVIDENCE_LOCATION_OVERRIDE_REJECTED: <supplied path> replaced with <canonical path>`.
+'''