prizmkit 1.1.127 → 1.1.129

package/bundled/dev-pipeline/scripts/generate-bootstrap-prompt.py

@@ -1192,8 +1192,8 @@ def load_active_agent_prompts(templates_dir):
     """Load explicit active agent prompt templates.
 
     Retired Dev implementation, review-fix, and plan-challenge prompts are
-    intentionally not loaded; feature implementation and accepted review fixes
-    are performed by the main orchestrator and reviewer gate.
+    intentionally not loaded; feature implementation, Code Review, and accepted
+    review fixes are performed directly by the current Main Agent.
     """
     agent_prompts_dir = os.path.join(templates_dir, "agent-prompts")
     if not os.path.isdir(agent_prompts_dir):
@@ -1235,11 +1235,12 @@ def _tier_header(pipeline_mode):
         ),
         "full": (
             "# Dev-Pipeline Session Bootstrap — Tier 3 "
-            "(Orchestrator + Reviewer)\n",
-            "**Tier 3 — Orchestrator + Reviewer**: For complex "
+            "(Main-Agent Full Guardrails)\n",
+            "**Tier 3 — Main-Agent Full Guardrails**: For complex "
             "features, the main orchestrator handles context, planning, "
-            "and implementation directly. Reviewer reviews, and no top-level "
-            "Dev implementation subagent is spawned.\n",
+            "implementation, and Code Review directly. No top-level Dev "
+            "implementation subagent is spawned, and Code Review never uses "
+            "direct or indirect delegation.\n",
         ),
     }
     return headers.get(pipeline_mode, headers["lite"])
@@ -1271,28 +1272,30 @@ def _tier_reminders(pipeline_mode):
         specific = [
             "- Tier 2: compatibility label only; feature prompts now render "
             "the unified full/tier3-equivalent guidance by default",
-            "- context-snapshot.md is append-only: orchestrator writes "
-            "Sections 1-4 + Implementation Log, Reviewer appends Review Notes",
-            "- Gate checks enforce Implementation Log (with Gate Evidence) "
-            "and Review Notes are written before proceeding",
-            "- Do NOT use `run_in_background=true` when spawning Reviewer",
-            "- On timeout: check snapshot + git diff HEAD → model:lite → "
-            "remaining steps only → max 2 retries per phase → "
-            "orchestrator fallback",
+            "- context-snapshot.md is the implementation knowledge base; "
+            "Code Review writes only review-report.md",
+            "- Gate checks require the Implementation Log (with Gate Evidence) "
+            "and a valid review-report.md Final Result before proceeding",
+            "- Code Review runs only in the current Main Agent and never uses "
+            "direct or indirect delegation",
+            "- On timeout outside Code Review: check snapshot + git diff HEAD "
+            "→ model:lite → remaining steps only → max 2 retries per phase → "
+            "Main Agent fallback",
         ]
     else:  # full
         specific = [
-            "- Tier 3: full orchestration — orchestrator implements "
-            "directly (no top-level Dev subagent), and Reviewer reviews",
-            "- context-snapshot.md is append-only: orchestrator writes "
-            "Sections 1-4 + Implementation Log, Reviewer appends "
-            "Review Notes",
-            "- Gate checks enforce Implementation Log and Review Notes are "
-            "written before proceeding",
+            "- Tier 3: full orchestration — the Main Agent implements directly "
+            "and owns the complete Code Review loop",
+            "- context-snapshot.md is the implementation knowledge base; "
+            "Code Review writes only review-report.md",
+            "- Gate checks require the Implementation Log and a valid "
+            "review-report.md Final Result before proceeding",
+            "- Code Review is excluded from normal work delegation and never "
+            "uses direct or indirect delegation or another review entry point",
             "- Do NOT use `run_in_background=true` when spawning normal work "
-            "agents",
-            "- If an Agent tool call fails with team/config/lock errors, retry "
-            "at most once; after a second failure, use the documented "
+            "agents outside Code Review",
+            "- If a normal work Agent call fails with team/config/lock errors, "
+            "retry at most once; after a second failure, use the documented "
             "inline/recovery fallback or write `failure-log.md` rather than "
             "spawning variants or polling indefinitely",
             "- NEVER delete, modify, or touch any file under `.prizmkit/state/` "
@@ -1550,9 +1553,8 @@ def validate_rendered(content):
     # contain literal double braces like Jinja or Go templates)
     unreplaced = re.findall(r"\{\{[A-Z][A-Z_0-9]+\}\}", content)
     if unreplaced:
-        # Deduplicate
         unique = sorted(set(unreplaced))
-        warnings.append(
+        errors.append(
             "Unreplaced placeholders: {}".format(", ".join(unique))
         )
 

package/bundled/dev-pipeline/scripts/generate-bugfix-prompt.py

@@ -356,7 +356,7 @@ def _bugfix_header(pipeline_mode):
     title = "# Dev-Pipeline Bug Fix Session Bootstrap — {}\n".format(pipeline_mode.title())
     desc = {
         "lite": "**Tier 1 — Single Agent**: direct root-cause fix by the main orchestrator.",
-        "standard": "**Tier 2 — Orchestrator + Reviewer**: direct implementation with review gates.",
+        "standard": "**Tier 2 — Main-Agent Review Gates**: direct implementation with Main-Agent-only review gates.",
         "full": "**Tier 3 — Full Bugfix Guardrails**: direct implementation with stronger diagnosis and review gates.",
     }.get(pipeline_mode, "**Bugfix pipeline**")
     return title + "\n" + desc + "\n"

package/bundled/dev-pipeline/scripts/generate-recovery-prompt.py

@@ -97,9 +97,10 @@ Implement the minimal fix (red → green):
         """\
 Verify fix quality:
 1. Run `/prizmkit-code-review` with the current bugfix artifact directory.
-2. The Main Agent self-reviews for up to ten rounds; semantic risk limits independent Reviewers to low=0, medium=1, high=2.
-3. Reviewer 2 requires an accepted high finding from Reviewer 1 plus repair, tests, and Main-Agent reconvergence; Reviewer 3 is forbidden.
-4. Continue only after the report's last Final Result is valid; preserve append-only progress for recovery.""",
+2. The current Main Agent is the only Code Review executor. It reviews the complete current change for up to ten rounds, directly repairs accepted findings, and verifies repairs.
+3. Converge only when accepted and unresolved findings are both zero; missing required evidence or round-ten non-convergence produces NEEDS_FIXES.
+4. Do not delegate review directly or indirectly. Do not invoke another review skill or review workflow. Do not relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep.
+5. `review-report.md` is the only persisted review artifact. Continue only after its last Final Result is valid; preserve append-only progress for recovery.""",
     ),
     6: (
         "User Verification",

package/bundled/dev-pipeline/scripts/generate-refactor-prompt.py

@@ -377,7 +377,7 @@ def _refactor_header(pipeline_mode):
     title = "# Dev-Pipeline Refactor Session Bootstrap — {}\n".format(pipeline_mode.title())
     desc = {
         "lite": "**Tier 1 — Single Agent**: direct behavior-preserving refactor by the main orchestrator.",
-        "standard": "**Tier 2 — Orchestrator + Reviewer**: direct refactor with review gates.",
+        "standard": "**Tier 2 — Main-Agent Review Gates**: direct refactor with Main-Agent-only review gates.",
         "full": "**Tier 3 — Full Refactor Guardrails**: direct refactor with stronger behavior-preservation evidence.",
     }.get(pipeline_mode, "**Refactor pipeline**")
     return title + "\n" + desc + "\n"

package/bundled/dev-pipeline/scripts/parse-stream-progress.py

@@ -977,8 +977,8 @@ class ProgressTracker:
             self.errors.append(normalized_code)
 
     def _record_agent_tool_call(self, provider, tool_name, raw_args, *, source_event="tool_use", event_id=""):
-        """Record Agent/TaskCreate tool calls as session-log subagent spawns."""
-        if tool_name not in ("Agent", "TaskCreate"):
+        """Record delegated execution tool calls as session-log subagent spawns."""
+        if tool_name != "Agent":
             return
         if isinstance(raw_args, dict):
             args = raw_args
@@ -1019,7 +1019,7 @@ class ProgressTracker:
         )
 
     def _record_cb_agent_tool_call(self, tool_name, raw_args, *, event_id="", source_event="tool_use"):
-        """Record a CodeBuddy Agent/TaskCreate tool invocation for sub-agent tracking."""
+        """Record a CodeBuddy delegated execution invocation for sub-agent tracking."""
         self._record_agent_tool_call(
             "codebuddy",
             tool_name,

package/bundled/dev-pipeline/templates/bootstrap-prompt.md

@@ -30,7 +30,7 @@ Infer what needs to be done from the feature context above and follow the standa
 
 3. **Implement**: Run `/prizmkit-implement` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to execute the plan using TDD (write tests first, then implement).
 
-4. **Review**: Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The Main Agent self-reviews for up to ten rounds; semantic risk then limits independent Reviewers to `low=0`, `medium=1`, `high=2`, with Reviewer 2 only after an accepted high finding is repaired and reconverged. Reviewer 3 is forbidden. Require the last `## Final Result` verdict before continuing.
+4. **Review**: Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The current Main Agent is the only Code Review executor. It reviews the complete current change for up to ten rounds, directly repairs accepted findings, verifies repairs, and converges when accepted and unresolved findings are both zero. Do not delegate review directly or indirectly, invoke another review skill/workflow, or relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep. `review-report.md` is the only persisted review artifact. Require the last `## Final Result` verdict before continuing.
 
 5. **Test**: Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to generate and verify tests only for this feature's changed scope after review. Do not use bugfix/refactor artifact directories for this gate.
 
@@ -44,7 +44,7 @@ Infer what needs to be done from the feature context above and follow the standa
 - Do NOT push to remote — the user will push manually.
 - Write all artifacts to `.prizmkit/specs/{{FEATURE_SLUG}}/`.
 - If a step fails after 3 attempts, write a status report and stop.
-- **Never re-read unmodified files**: once you read a file into context, do NOT read it again unless it was modified by an Edit/Write after your last read — use `grep -n`/`sed -n` for specific lines. Before issuing a Read, ask: "Have I read this file before, and has it changed since?" If no — do NOT read it. Reviewer subagents must rely on context-snapshot.md rather than re-reading files the orchestrator already read.
+- **Never re-read unmodified files**: once you read a file into context, do NOT read it again unless it was modified by an Edit/Write after your last read — use `grep -n`/`sed -n` for specific lines. Before issuing a Read, ask: "Have I read this file before, and has it changed since?" If no — do NOT read it. Normal work delegation outside Code Review must rely on context-snapshot.md rather than re-reading files the Main Agent already read. Code Review runs only in the current Main Agent.
 - **Read offset safety**: if Read returns "shorter than offset"/"empty" — run `wc -l` then `sed -n`, do NOT retry same offset.
 - **Hard Round Cap**: after 6 total test-fix rounds, STOP trial-and-error. Switch to read-once-then-rewrite. Do NOT exceed 7 rounds.
 - **Layered test**: debug with single-file tests, only run full suite as gate when single-file is green.

package/bundled/dev-pipeline/templates/bootstrap-tier3.md

@@ -1,4 +1,4 @@
-# Dev-Pipeline Session Bootstrap — Tier 3 (Orchestrator + Reviewer)
+# Dev-Pipeline Session Bootstrap — Tier 3 (Main-Agent Full Guardrails)
 
 ## Session Context
 
@@ -10,12 +10,12 @@
 
 You are the **session orchestrator**. Implement Feature {{FEATURE_ID}}: "{{FEATURE_TITLE}}".
 
-**CRITICAL**: You MUST NOT exit until ALL work is complete and committed. When you spawn Reviewer agents, wait for each to finish (run_in_background=false). Do NOT spawn agents in background and exit — that kills the session.
+**CRITICAL**: You MUST NOT exit until ALL work is complete and committed. Normal work delegation outside Code Review must finish before exit (`run_in_background=false`). Code Review never uses direct or indirect delegation.
 
-**Tier 3 — Orchestrator + Reviewer**: For complex features, the main orchestrator handles context, planning, and implementation directly. Do not spawn top-level Dev or Reviewer subagents; code review runs later through `/prizmkit-code-review`.
+**Tier 3 — Main-Agent Full Guardrails**: For complex features, the Main Agent handles context, planning, implementation, and Code Review directly. Do not spawn a top-level Dev implementation subagent. Code Review runs through `/prizmkit-code-review` without direct or indirect delegation.
 
 **Agent spawn failure policy (all Agent tool calls)**:
-- If spawning Reviewer fails with team/config/lock errors, retry at most once.
+- Code Review must not invoke another review skill/workflow or relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep.
 - If the second attempt fails, do not keep spawning variants and do not enter artifact polling for Implementation Log, challenge report, or review report markers.
 - Use the documented inline/recovery fallback for that phase: write the required report yourself where possible, complete remaining work directly in the orchestrator when safe, or write `failure-log.md` with the spawn error and last observable state before stopping for recovery.
 - Apply the same cap to any re-spawn for report repair or resume prompts; do not burn multiple minutes on identical team/config/lock failures.
@@ -52,7 +52,7 @@ You are running in **headless non-interactive mode** with a FINITE context windo
 
 1. **context-snapshot.md is your single source of truth** — After Phase 1-2 makes it available, use it instead of re-reading individual source files. If you just created or updated it in this session, continue from the in-context content; read it only when resuming/continuing, after context compaction/loss, or when handing work to a subagent.
 2. **Never re-read your own writes** — After you create/modify a file, do NOT read it back to verify. Trust your write was correct.
-3. **Never re-read unmodified files** — Once you have read a file into context, do NOT read it again unless it was modified by an Edit/Write operation after your last read. Use platform file/search tools for targeted lookup; when shell-free helper lookup is needed, run `{{RUNTIME_HELPER_CMD}} text search <file> --pattern "<pattern>" --json`, then read only the smallest bounded range with the platform file-reading tool instead of full-file Read. Before issuing a Read, ask yourself: "Have I read this file before, and has it changed since?" If the answer is "no, it hasn't changed" — do NOT read it. This applies to all files including those summarized in context-snapshot.md Section 4. Reviewer subagents must receive context-snapshot.md and rely on it rather than independently re-reading the same files the orchestrator already read.
+3. **Never re-read unmodified files** — Once you have read a file into context, do NOT read it again unless it was modified by an Edit/Write operation after your last read. Use platform file/search tools for targeted lookup; when shell-free helper lookup is needed, run `{{RUNTIME_HELPER_CMD}} text search <file> --pattern "<pattern>" --json`, then read only the smallest bounded range with the platform file-reading tool instead of full-file Read. Before issuing a Read, ask yourself: "Have I read this file before, and has it changed since?" If the answer is "no, it hasn't changed" — do NOT read it. This applies to all files including those summarized in context-snapshot.md Section 4. Normal work delegation outside Code Review must receive context-snapshot.md and rely on it rather than independently re-reading files the Main Agent already read. Code Review runs only in the current Main Agent.
 4. **Stay focused** — Do NOT explore code unrelated to this feature. No curiosity-driven reads.
 4. **One task at a time** — In Phase 4 (implement), complete and test one task before starting the next.
 5. **Minimize tool output** — Never load full command output into context. First capture command output to a temp file using the command/tool output option or the host-appropriate redirection, then inspect only the final relevant lines and use targeted search helpers to extract the information needed for the current task. Only read the filtered result — never the raw full output.
@@ -67,12 +67,12 @@ You are running in **headless non-interactive mode** with a FINITE context windo
 ## PrizmKit Directory Convention
 
 ```
-.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md  ← orchestrator writes Sections 1-4 + Implementation Log; Reviewer appends Review Notes
+.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md  ← Main Agent writes Sections 1-4 + Implementation Log
 .prizmkit/specs/{{FEATURE_SLUG}}/spec.md
 .prizmkit/specs/{{FEATURE_SLUG}}/plan.md              ← includes Tasks section
 ```
 
-**`context-snapshot.md`** is the shared knowledge base. Orchestrator writes Sections 1-4 and the Implementation Log; Reviewer appends Review Notes. This eliminates redundant I/O across review and orchestration phases.
+**`context-snapshot.md`** is the implementation knowledge base. The Main Agent writes Sections 1-4 and the Implementation Log. Code Review writes only `review-report.md` and never appends separate Review Notes.
 
 ---
 
@@ -80,7 +80,7 @@ You are running in **headless non-interactive mode** with a FINITE context windo
 
 If any agent times out:
 1. `{{RUNTIME_HELPER_CMD}} artifact list .prizmkit/specs/{{FEATURE_SLUG}}` — check what exists
-2. If `context-snapshot.md` exists: open recovery prompt with `"Read .prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md for project context and any Implementation Log/Review Notes from previous agents. Run git diff HEAD to see actual code changes already made. Do NOT re-read individual source files unless the File Manifest directs you to."` + only remaining steps + `model: "lite"`
+2. If `context-snapshot.md` exists: open recovery prompt with `"Read .prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md for project context and the Implementation Log from previous normal work execution. Run git diff HEAD to see actual code changes already made. Do NOT re-read individual source files unless the File Manifest directs you to."` + only remaining steps + `model: "lite"`
 3. Max 2 retries per phase. After 2 failures, orchestrator completes the work directly and appends a Recovery Note to context-snapshot.md.
 
 ---
@@ -90,22 +90,21 @@ If any agent times out:
 {{IF_INIT_NEEDED}}
 ### Phase 0: Project Bootstrap
 - Run `/prizmkit-init` (invoke the prizmkit-init skill)
-- Run `{{PYTHON_CMD}} {{INIT_SCRIPT_PATH}} --project-root {{PROJECT_ROOT}} --feature-id {{FEATURE_ID}} --feature-slug {{FEATURE_SLUG}}`
+- Run `{{PYTHON_CMD}} .prizmkit/dev-pipeline/scripts/init-change-artifact.py --project-root {{PROJECT_ROOT}} --feature-id {{FEATURE_ID}} --feature-slug {{FEATURE_SLUG}}`
 - **CP-0**: Verify `.prizmkit/prizm-docs/root.prizm`, `.prizmkit/config.json` exist
 {{END_IF_INIT_NEEDED}}
-{{IF_INIT_DONE}}
 
 ### Step 1: Initialize
 
 1. Run init script:
-   `{{PYTHON_CMD}} {{INIT_SCRIPT_PATH}} --project-root {{PROJECT_ROOT}} --feature-id {{FEATURE_ID}} --feature-slug {{FEATURE_SLUG}}`
+   `{{PYTHON_CMD}} .prizmkit/dev-pipeline/scripts/init-change-artifact.py --project-root {{PROJECT_ROOT}} --feature-id {{FEATURE_ID}} --feature-slug {{FEATURE_SLUG}}`
 
 2. Check for existing artifacts:
    `{{RUNTIME_HELPER_CMD}} artifact exists .prizmkit/specs/{{FEATURE_SLUG}}`
 
 {{IF_FRESH_START}}
 ```bash
-{{PYTHON_CMD}} {{INIT_SCRIPT_PATH}} --project-root {{PROJECT_ROOT}} --feature-id {{FEATURE_ID}} --feature-slug {{FEATURE_SLUG}}
+{{PYTHON_CMD}} .prizmkit/dev-pipeline/scripts/init-change-artifact.py --project-root {{PROJECT_ROOT}} --feature-id {{FEATURE_ID}} --feature-slug {{FEATURE_SLUG}}
 ```
 {{END_IF_FRESH_START}}
 
@@ -182,7 +181,7 @@ Before proceeding past CP-1, verify:
 
 ### Phase 3: Plan Self-Check — Orchestrator Direct
 
-Do not spawn a top-level Reviewer subagent for planning analysis. Perform a bounded self-check before implementation:
+Do not delegate planning analysis. The current Main Agent performs a bounded self-check before implementation:
 
 1. Re-read only `context-snapshot.md`, `spec.md`, and `plan.md`.
 2. Cross-check spec.md and plan.md Tasks against the feature description and acceptance criteria.
@@ -208,7 +207,7 @@ Before starting, check plan.md Tasks:
 **3a.** Run `/prizmkit-implement` directly:
 - Continue from the `spec.md` / `plan.md` / `context-snapshot.md` content you created in Phase 1-2; do not re-read self-authored artifacts just because implementation starts
 - Re-read `plan.md` or `context-snapshot.md` only when resuming/continuing, when current context was compacted or lost, or when task checkbox state may have changed outside this session
-- If implementation is delegated to a Dev subagent, the subagent must read `spec.md` if the artifact exists, plus `plan.md` and `context-snapshot.md`, before editing
+- Implementation remains in the current Main Agent; do not delegate it to a Dev subagent
 - Use context-snapshot.md Section 4 File Manifest for targeted source reads; avoid re-reading source files already summarized there unless they changed after the snapshot
 - Implements task-by-task, marking each `[x]` immediately
 - Remeber to update checkpoints after finishing a series of tasks
@@ -238,15 +237,17 @@ For each Verification Gate from the Task Contract, write one evidence line to Im
 
 ### Review — Code Review
 
-Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so the code-review skill can read spec, plan, context snapshot, implementation log, and current diff artifacts. Do not require or read a prior `/prizmkit-test` report here; test generation and test-fix looping happen in the next phase.
+Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so `/prizmkit-code-review` can read the spec, plan, context snapshot, implementation log, and current diff. Do not require or read a prior `/prizmkit-test` report here; test generation and test-fix looping happen in the next phase.
 
-Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The skill first runs a Main-Agent self-review loop for up to ten completed rounds, then semantic risk permits independent Reviewers at limits `low=0`, `medium=1`, `high=2`. High risk starts Reviewer 2 only after an accepted high finding was repaired, tested, and reconverged; Reviewer 3 is forbidden. The skill owns launch parameters, current-checkout preference, forced-worktree changed-content fallback, and append-only report progress.
+Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The skill runs a Main-Agent review loop for up to ten completed rounds. The current Main Agent is the only Code Review executor: it directly reviews, repairs accepted findings, verifies repairs, and converges when accepted and unresolved findings are both zero.
+
+The Code Review execution boundary is absolute: do not delegate review directly or indirectly. Do not invoke another review skill or review workflow. Do not relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep. Direct reading, searching, editing, and testing by the current Main Agent remain allowed. `review-report.md` is the only persisted review artifact. Missing required evidence, unsafe repair, failed verification, or round-ten non-convergence produces `NEEDS_FIXES`.
 
 **Gate Check — Final Review Result**:
 ```bash
 {{RUNTIME_HELPER_CMD}} text final-verdict .prizmkit/specs/{{FEATURE_SLUG}}/review-report.md
 ```
-Handle `REVIEW_PASS`, `REVIEW_NEEDS_FIXES`, and `REVIEW_BLOCKED` according to task policy. `REVIEW_INCOMPLETE` or `REVIEW_INVALID` means review did not complete: inspect progress once, write `failure-log.md`, and stop. Do not accept old or incidental verdict text or synthesize a passing result through Main-Agent self-review.
+Handle `REVIEW_PASS` and `REVIEW_NEEDS_FIXES` according to task policy. `REVIEW_INCOMPLETE` or `REVIEW_INVALID` means review did not complete: inspect progress once, write `failure-log.md`, and stop. Do not accept old or incidental verdict text or synthesize a passing result through Main-Agent self-review.
 
 **CP-3**: Review has a valid Final Result and the report preserves this execution's progress.
 
@@ -266,7 +267,7 @@ Gate requirements:
 - Record unrelated historical gaps under `Out of Scope Gaps`; do not generate tests for them and do not fail this feature because of them.
 - Accept only a current-run `.prizmkit/test/*/test-report.md` whose `Scope` mode is `this-change`, whose `Artifact Dir` matches `.prizmkit/specs/{{FEATURE_SLUG}}/`, whose `Verdict` is `PASS`, whose `Boundary Completion Gate` reports `Completion gate passed: yes` with zero `Boundary-missing` and zero `Happy-path-only`, and whose `Boundary Validation` result is `passed`.
 - Only after a valid `PASS`, write the report path to `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt` and append a 3-5 bullet `## PrizmKit Test Gate` summary to `context-snapshot.md`.
-- If the report verdict is `NEEDS_FIXES`, fix in-scope implementation/test issues and rerun `/prizmkit-test`. If the report is `BLOCKED`, missing, stale, or for the wrong scope/artifact dir, write `failure-log.md` and stop for recovery.
+- If the report verdict is `NEEDS_FIXES`, fix in-scope implementation/test issues and rerun `/prizmkit-test`. If the report is missing, stale, invalid, or for the wrong scope/artifact dir, write `failure-log.md` and stop for recovery.
 
 {{IF_BROWSER_INTERACTION}}
 ### Phase 5.5: Browser Verification — MANDATORY
@@ -518,9 +519,9 @@ If you encounter an unrecoverable error, context overflow, or are about to exit
 ## Reminders
 
 - Tier 3: orchestrator implements directly; `/prizmkit-code-review` handles review later
-- context-snapshot.md is append-only: orchestrator writes Sections 1-4 + Implementation Log, Reviewer appends Review Notes
-- Gate checks enforce Implementation Log and Review Notes are written before proceeding
-- Do NOT use `run_in_background=true` when spawning Reviewer agents
+- context-snapshot.md contains Main-Agent implementation context; Code Review writes only review-report.md
+- Gate checks require the Implementation Log and a valid review-report.md Final Result before proceeding
+- Code Review never uses direct or indirect delegation; normal work delegation outside review must not use `run_in_background=true`
 - Commit phase must use `/prizmkit-committer`; do NOT replace with manual git commit commands
 - **NEVER delete, modify, or touch any file under `.prizmkit/state/`** — those are pipeline runtime files managed by the runner
 - NEVER run `rm -rf`, `git clean`, or any destructive filesystem operations

package/bundled/dev-pipeline/templates/bugfix-bootstrap-prompt.md

@@ -180,15 +180,15 @@ After implement completes, verify:
 If `FAST_PATH=true` (≤ 2 tasks, obvious root cause), skip this phase entirely and mark checkpoint step `prizmkit-code-review` as `skipped` before continuing.
 
 Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/bugfix/{{BUG_ID}}/`:
-- The skill first runs up to ten Main-Agent self-review rounds, then semantic risk limits independent Reviewers to `low=0`, `medium=1`, `high=2`
-- High risk starts Reviewer 2 only after an accepted high finding was repaired, tested, and reconverged; Reviewer 3 is forbidden
-- The skill preserves its independent Reviewer prompt and appends progress to `review-report.md`
+- The skill runs up to ten Main-Agent review rounds, directly repairs accepted findings, and verifies each repair
+- Missing required evidence, unsafe repair, failed verification, or round-ten non-convergence produces `NEEDS_FIXES`
+- The skill appends Main-Agent review and repair progress to `review-report.md`
 
 **Gate Check — Final Review Result**:
 ```bash
 {{RUNTIME_HELPER_CMD}} text final-verdict .prizmkit/bugfix/{{BUG_ID}}/review-report.md
 ```
-Handle `REVIEW_PASS`, `REVIEW_NEEDS_FIXES`, and `REVIEW_BLOCKED` according to task policy. `REVIEW_INCOMPLETE` or `REVIEW_INVALID` means review did not complete: inspect progress once, write `failure-log.md`, and stop. Do not accept stale verdict text or replace required independent review with Main-Agent self-review.
+Handle `REVIEW_PASS` and `REVIEW_NEEDS_FIXES` according to task policy. `REVIEW_INCOMPLETE` or `REVIEW_INVALID` means review did not complete: inspect progress once, write `failure-log.md`, and stop. Do not accept stale verdict text or replace required independent review with Main-Agent self-review.
 
 {{IF_BROWSER_INTERACTION}}
 

package/bundled/dev-pipeline/templates/refactor-bootstrap-prompt.md

@@ -17,7 +17,7 @@ You are the **refactor session orchestrator**. Execute Refactor {{REFACTOR_ID}}:
 
 **NON-INTERACTIVE MODE**: You are running in headless non-interactive mode. There is NO human on the other end. NEVER ask for user confirmation, NEVER wait for user input, and NEVER use interactive prompts. Resolve uncertainty from existing code, project docs, and conservative behavior-preserving defaults.
 
-**ORCHESTRATOR-DIRECT RULE**: Do not spawn top-level Dev or Reviewer subagents. The main orchestrator plans, implements via `/prizmkit-implement`, and runs `/prizmkit-code-review` directly.
+**MAIN-AGENT-DIRECT RULE**: Do not spawn a top-level Dev implementation subagent. The Main Agent plans, implements via `/prizmkit-implement`, and runs `/prizmkit-code-review` directly without direct or indirect review delegation.
 
 **REFACTOR DOCUMENTATION POLICY**:
 - Default: run `/prizmkit-retrospective` with full sync because refactoring changes structure and interfaces.
@@ -139,7 +139,7 @@ Apply the browser behavior preservation protocol where UI behavior can be affect
 
 ### Phase 4: Review — Code Review & Behavior Verification
 
-Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/refactor/{{REFACTOR_ID}}/` directly from the main orchestrator. The skill first runs up to ten Main-Agent self-review rounds, then semantic risk limits independent Reviewers to `low=0`, `medium=1`, `high=2`. High risk starts Reviewer 2 only after an accepted high finding was repaired, tested, and reconverged; Reviewer 3 is forbidden.
+Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/refactor/{{REFACTOR_ID}}/` directly from the main orchestrator. The skill runs up to ten Main-Agent review rounds, directly repairs accepted findings, and verifies each repair. Missing required evidence, unsafe repair, failed verification, or round-ten non-convergence produces `NEEDS_FIXES`.
 
 Review requirements:
 - verify refactor quality against spec.md and plan.md;
@@ -151,9 +151,9 @@ Review requirements:
 ```bash
 {{RUNTIME_HELPER_CMD}} text final-verdict .prizmkit/refactor/{{REFACTOR_ID}}/review-report.md
 ```
-Handle `REVIEW_PASS`, `REVIEW_NEEDS_FIXES`, and `REVIEW_BLOCKED` according to behavior-preservation policy. `REVIEW_INCOMPLETE` or `REVIEW_INVALID` means review did not complete: inspect progress once, write `failure-log.md`, and stop. Do not accept stale verdict text or replace required independent review with Main-Agent self-review.
+Handle `REVIEW_PASS` and `REVIEW_NEEDS_FIXES` according to behavior-preservation policy. `REVIEW_INCOMPLETE` or `REVIEW_INVALID` means review did not complete: inspect progress once, write `failure-log.md`, and stop. Do not accept stale verdict text or replace required independent review with Main-Agent self-review.
 
-**Checkpoint update**: set step `prizmkit-code-review` to `completed` only for an accepted valid Final Result. Do not complete it for blocked, incomplete, or invalid reports.
+**Checkpoint update**: set step `prizmkit-code-review` to `completed` only for an accepted valid Final Result. Do not complete it for incomplete or invalid reports.
 
 ---
 
@@ -249,7 +249,7 @@ Before exiting, write `{{SESSION_STATUS_PATH}}`:
 ## Reminders
 
 - Use L1 skills: `/prizmkit-plan`, `/prizmkit-implement`, `/prizmkit-code-review`, `/prizmkit-retrospective`, `/prizmkit-committer`.
-- Do not spawn top-level Dev or Reviewer subagents.
+- Do not spawn a top-level Dev implementation subagent; Code Review runs only in the current Main Agent without delegation.
 - Behavior preservation is the first priority.
 - Browser verification is a mandatory attempt by default; it blocks only when explicitly required.
 - If `DEV_PORT` is unknown, discover it from project config; do not guess port 3000.

package/bundled/dev-pipeline/templates/sections/bugfix-mission.md

@@ -6,4 +6,4 @@ You are the **bug fix session orchestrator**. Fix Bug {{BUG_ID}}: "{{BUG_TITLE}}
 
 **NON-INTERACTIVE MODE**: There is NO human on the other end. NEVER ask for user confirmation, NEVER wait for user input. Make decisions autonomously from the bug report, existing code, tests, and conservative project patterns.
 
-**Execution model**: You implement directly in the current checkout. Do not spawn a top-level Dev subagent. `/prizmkit-code-review` owns its internal Reviewer Agent → filter → orchestrator fix loop.
+**Execution model**: You implement directly in the current checkout. Do not spawn a top-level Dev subagent. The current Main Agent runs `/prizmkit-code-review` directly and owns its complete review, repair, verification, and reporting loop without direct or indirect review delegation.

package/bundled/dev-pipeline/templates/sections/bugfix-phase-review.md

@@ -12,9 +12,17 @@ If `FAST_PATH=true`, skip this phase intentionally:
 
 Then continue to the next checkpoint step.
 
-Otherwise, run `/prizmkit-code-review` with `artifact_dir=.prizmkit/bugfix/{{BUG_ID}}/`. The skill first runs a Main-Agent self-review loop for up to ten completed rounds, then semantic risk permits independent Reviewers at limits `low=0`, `medium=1`, `high=2`. High risk starts Reviewer 2 only after Reviewer 1 produced an accepted high finding that was repaired, tested, and reconverged. Reviewer 3 is forbidden.
+Otherwise, run `/prizmkit-code-review` with `artifact_dir=.prizmkit/bugfix/{{BUG_ID}}/`. The skill runs a Main-Agent review loop for up to ten completed rounds. The current Main Agent is the only Code Review executor: it reviews the complete current change, rejects unsupported candidate findings, directly repairs accepted findings, verifies repairs, and repeats until `accepted = 0` with no unresolved findings.
 
-The skill owns its unchanged independent Reviewer prompt, current-checkout preference, forced-worktree changed-content fallback, one bounded infrastructure retry per required Reviewer position, and append-only progress in `review-report.md`.
+The Code Review execution boundary is absolute for this phase:
+
+- Do not delegate any review work directly or indirectly.
+- Do not invoke another review skill or review workflow.
+- Do not use a general-purpose execution unit for review or relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep.
+- Direct reading, searching, editing, and testing by the current Main Agent remain allowed.
+- `review-report.md` is the only persisted review artifact; do not append separate Review Notes to another artifact.
+
+Missing required evidence, an unsafe repair, a failed verification, or failure to converge by round ten produces `NEEDS_FIXES`.
 
 **Gate Check — Final Review Result**:
 
@@ -24,7 +32,6 @@ The skill owns its unchanged independent Reviewer prompt, current-checkout prefe
 
 - `REVIEW_PASS` → proceed.
 - `REVIEW_NEEDS_FIXES` → log unresolved findings and proceed only when explicit task policy allows them.
-- `REVIEW_BLOCKED` → log infrastructure evidence and stop so recovery can retry.
-- `REVIEW_INCOMPLETE` or `REVIEW_INVALID` → inspect progress once, write `failure-log.md`, and stop. Do not accept old or incidental verdict text and do not replace required independent review with Main-Agent self-review.
+- `REVIEW_INCOMPLETE` or `REVIEW_INVALID` → inspect progress once, write `failure-log.md`, and stop. Do not accept old or incidental verdict text and do not synthesize a passing result outside the skill.
 
 **Checkpoint update**: Set step `prizmkit-code-review` to `completed` only for an accepted valid Final Result, or `skipped` for FAST_PATH.

package/bundled/dev-pipeline/templates/sections/context-budget-rules.md

@@ -6,7 +6,7 @@ You are running in **headless non-interactive mode** with a FINITE context windo
 
 1. **context-snapshot.md is your single source of truth** — After it is available, use it instead of re-reading individual source files. If you just created or updated it in this session, continue from the in-context content; read it only when resuming/continuing, after context compaction/loss, or when handing work to a subagent.
 2. **Never re-read your own writes** — After you create/modify a file, do NOT read it back to verify. Trust your write was correct.
-3. **Never re-read unmodified files** — Once you have read a file into context, do NOT read it again unless it was modified by an Edit/Write operation after your last read. Use platform file/search tools for targeted lookup; when shell-free helper lookup is needed, run `{{RUNTIME_HELPER_CMD}} text search <file> --pattern "<pattern>" --json`, then read only the smallest bounded range with the platform file-reading tool instead of full-file Read. Before issuing a Read, ask yourself: "Have I read this file before, and has it changed since?" If the answer is "no, it hasn't changed" — do NOT read it. This applies to all files including those summarized in context-snapshot.md Section 4. Reviewer subagents must receive context-snapshot.md and rely on it rather than independently re-reading the same files the orchestrator already read.
+3. **Never re-read unmodified files** — Once you have read a file into context, do NOT read it again unless it was modified by an Edit/Write operation after your last read. Use platform file/search tools for targeted lookup; when shell-free helper lookup is needed, run `{{RUNTIME_HELPER_CMD}} text search <file> --pattern "<pattern>" --json`, then read only the smallest bounded range with the platform file-reading tool instead of full-file Read. Before issuing a Read, ask yourself: "Have I read this file before, and has it changed since?" If the answer is "no, it hasn't changed" — do NOT read it. This applies to all files summarized in context-snapshot.md Section 4. Normal work delegation outside Code Review must receive context-snapshot.md and rely on it rather than independently re-reading files the Main Agent already read. Code Review is excluded from normal work delegation and runs only in the current Main Agent.
 4. **Stay focused** — Do NOT explore code unrelated to this feature. No curiosity-driven reads.
 5. **One task at a time** — Complete and test one task before starting the next.
 6. **Minimize tool output** — Never load full command output into context. First capture command output to a temp file using the command/tool output option or the host-appropriate redirection, then inspect only the final relevant lines and use targeted search helpers to extract the information needed for the current task. Only read the filtered result — never the raw full output.

package/bundled/dev-pipeline/templates/sections/directory-convention-agent.md

@@ -1,8 +1,9 @@
 ## PrizmKit Directory Convention
 
 ```
-.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md  ← orchestrator writes Sections 1-4 + Implementation Log; Reviewer appends Review Notes
+.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md  ← Main Agent writes Sections 1-4 + Implementation Log
 .prizmkit/specs/{{FEATURE_SLUG}}/plan.md              ← includes Tasks section
+.prizmkit/specs/{{FEATURE_SLUG}}/review-report.md     ← Main Agent writes Code Review progress and Final Result
 ```
 
-**`context-snapshot.md`** is the shared knowledge base. Orchestrator writes Sections 1-4 and the Implementation Log; Reviewer appends Review Notes. Append-only after initial creation.
+**`context-snapshot.md`** is the implementation knowledge base and remains append-only after initial creation. Code Review writes only `review-report.md`; it does not append separate Review Notes to the context snapshot.

package/bundled/dev-pipeline/templates/sections/directory-convention-full.md

@@ -1,9 +1,10 @@
 ## PrizmKit Directory Convention
 
 ```
-.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md  ← orchestrator writes Sections 1-4 + Implementation Log; Reviewer appends Review Notes
+.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md  ← Main Agent writes Sections 1-4 + Implementation Log
 .prizmkit/specs/{{FEATURE_SLUG}}/spec.md
 .prizmkit/specs/{{FEATURE_SLUG}}/plan.md              ← includes Tasks section
+.prizmkit/specs/{{FEATURE_SLUG}}/review-report.md     ← Main Agent writes Code Review progress and Final Result
 ```
 
-**`context-snapshot.md`** is the shared knowledge base. Orchestrator writes Sections 1-4 and the Implementation Log; Reviewer appends Review Notes. This eliminates redundant I/O across review and orchestration phases.
+**`context-snapshot.md`** is the implementation knowledge base. The Main Agent writes Sections 1-4 and the Implementation Log. Code Review writes only `review-report.md`; it does not append separate Review Notes to the context snapshot.

package/bundled/dev-pipeline/templates/sections/phase-commit-full.md

@@ -72,4 +72,4 @@ After all of the following are true, the session is **TERMINAL**:
 - `.prizmkit/specs/{{FEATURE_SLUG}}/completion-summary.json` is written
 - the `prizmkit-committer` checkpoint is marked `completed`
 
-At that point, stop immediately. Do **not** process delayed teammate/reviewer findings, do **not** edit files, do **not** amend the commit, and do **not** start any new tool calls. Any findings that arrive after S09 is completed are follow-up work for a later session, not part of this feature commit.
+At that point, stop immediately. Do **not** process delayed normal-work results, edit files, amend the commit, or start any new tool calls. Code Review cannot have delayed delegated findings because it runs only in the current Main Agent. Any normal-work result that arrives after S09 is completed is follow-up work for a later session, not part of this feature commit.

package/bundled/dev-pipeline/templates/sections/phase-commit.md

@@ -58,4 +58,4 @@ After all of the following are true, the session is **TERMINAL**:
 - `.prizmkit/specs/{{FEATURE_SLUG}}/completion-summary.json` is written
 - the `prizmkit-committer` checkpoint is marked `completed`
 
-At that point, stop immediately. Do **not** process delayed teammate/reviewer findings, do **not** edit files, do **not** amend the commit, and do **not** start any new tool calls. Any findings that arrive after S09 is completed are follow-up work for a later session, not part of this feature commit.
+At that point, stop immediately. Do **not** process delayed normal-work results, edit files, amend the commit, or start any new tool calls. Code Review cannot have delayed delegated findings because it runs only in the current Main Agent. Any normal-work result that arrives after S09 is completed is follow-up work for a later session, not part of this feature commit.

package/bundled/dev-pipeline/templates/sections/phase-implement-agent.md

@@ -14,7 +14,7 @@ Before starting, check plan.md Tasks:
 **3a.** Run `/prizmkit-implement` directly:
 - Continue from the `spec.md` / `plan.md` / `context-snapshot.md` content you created in Phase 1-2; do not re-read self-authored artifacts just because implementation starts
 - Re-read `plan.md` or `context-snapshot.md` only when resuming/continuing, when current context was compacted or lost, or when task checkbox state may have changed outside this session
-- If implementation is delegated to a Dev subagent, the subagent must read `spec.md` if the artifact exists, plus `plan.md` and `context-snapshot.md`, before editing
+- Implementation remains in the current Main Agent; do not delegate it to a Dev subagent
 - Use context-snapshot.md Section 4 File Manifest for targeted source reads; avoid re-reading source files already summarized there unless they changed after the snapshot
 - Implements task-by-task, marking each `[x]` immediately
 - Remeber to update checkpoints after finishing a series of tasks

package/bundled/dev-pipeline/templates/sections/phase-implement-full.md

@@ -14,7 +14,7 @@ Before starting, check plan.md Tasks:
 **3a.** Run `/prizmkit-implement` directly:
 - Continue from the `spec.md` / `plan.md` / `context-snapshot.md` content you created in Phase 1-2; do not re-read self-authored artifacts just because implementation starts
 - Re-read `plan.md` or `context-snapshot.md` only when resuming/continuing, when current context was compacted or lost, or when task checkbox state may have changed outside this session
-- If implementation is delegated to a Dev subagent, the subagent must read `spec.md` if the artifact exists, plus `plan.md` and `context-snapshot.md`, before editing
+- Implementation remains in the current Main Agent; do not delegate it to a Dev subagent
 - Use context-snapshot.md Section 4 File Manifest for targeted source reads; avoid re-reading source files already summarized there unless they changed after the snapshot
 - Implements task-by-task, marking each `[x]` immediately
 - Remeber to update checkpoints after finishing a series of tasks

package/bundled/dev-pipeline/templates/sections/phase-review-agent.md

@@ -1,10 +1,18 @@
 ### Review — Code Review
 
-Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so the code-review skill can read spec, plan, context snapshot, implementation log, and current diff artifacts. Do not require a prior `/prizmkit-test` report here; the next phase owns the formal test gate.
+Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so `/prizmkit-code-review` can read the spec, plan, context snapshot, implementation log, and current diff. Do not require a prior `/prizmkit-test` report here; the next phase owns the formal test gate.
 
-Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The skill first runs a Main-Agent self-review loop for up to ten completed rounds, then semantic risk permits independent Reviewers at limits `low=0`, `medium=1`, `high=2`. High risk starts Reviewer 2 only after Reviewer 1 produced an accepted high finding that was repaired, tested, and reconverged. Reviewer 3 is forbidden.
+Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The skill runs a Main-Agent review loop for up to ten completed rounds. The current Main Agent is the only Code Review executor: it reviews the complete current change, rejects unsupported candidate findings, directly repairs accepted findings, verifies repairs, and repeats until `accepted = 0` with no unresolved findings.
 
-The skill owns Reviewer launch parameters, current-checkout preference, forced-worktree changed-content fallback, one bounded infrastructure retry per required Reviewer position, and append-only progress in `review-report.md`.
+The Code Review execution boundary is absolute for this phase:
+
+- Do not delegate any review work directly or indirectly.
+- Do not invoke another review skill or review workflow.
+- Do not use a general-purpose execution unit for review or relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep.
+- Direct reading, searching, editing, and testing by the current Main Agent remain allowed.
+- `review-report.md` is the only persisted review artifact; do not append separate Review Notes to `context-snapshot.md`.
+
+Missing required evidence, an unsafe repair, a failed verification, or failure to converge by round ten produces `NEEDS_FIXES`.
 
 **Gate Check — Final Review Result**:
 
@@ -16,8 +24,7 @@ Handle the marker:
 
 - `REVIEW_PASS` → proceed to next phase.
 - `REVIEW_NEEDS_FIXES` → log unresolved findings and follow the task failure policy; do not externally restart the review loop.
-- `REVIEW_BLOCKED` → log infrastructure evidence and stop so recovery can retry.
-- `REVIEW_INCOMPLETE` or `REVIEW_INVALID` → inspect the append-only progress once, write `failure-log.md`, and stop. Do not treat an old or incidental verdict as completion and do not synthesize a passing result through Main-Agent self-review.
+- `REVIEW_INCOMPLETE` or `REVIEW_INVALID` → inspect the append-only progress once, write `failure-log.md`, and stop. Do not treat an old or incidental verdict as completion and do not synthesize a passing result outside the skill.
 
 **CP-3**: Review has a valid Final Result and `review-report.md` preserves this execution's progress.
 

package/bundled/dev-pipeline/templates/sections/phase-review-full.md

@@ -1,10 +1,18 @@
 ### Review — Code Review
 
-Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so the code-review skill can read spec, plan, context snapshot, implementation log, and current diff artifacts. Do not require a prior `/prizmkit-test` report here; the next phase owns the formal test gate.
+Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so `/prizmkit-code-review` can read the spec, plan, context snapshot, implementation log, and current diff. Do not require a prior `/prizmkit-test` report here; the next phase owns the formal test gate.
 
-Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The skill first runs a Main-Agent self-review loop for up to ten completed rounds, then semantic risk permits independent Reviewers at limits `low=0`, `medium=1`, `high=2`. High risk starts Reviewer 2 only after Reviewer 1 produced an accepted high finding that was repaired, tested, and reconverged. Reviewer 3 is forbidden.
+Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`. The skill runs a Main-Agent review loop for up to ten completed rounds. The current Main Agent is the only Code Review executor: it reviews the complete current change, rejects unsupported candidate findings, directly repairs accepted findings, verifies repairs, and repeats until `accepted = 0` with no unresolved findings.
 
-The skill owns Reviewer launch parameters, current-checkout preference, forced-worktree changed-content fallback, one bounded infrastructure retry per required Reviewer position, and append-only progress in `review-report.md`.
+The Code Review execution boundary is absolute for this phase:
+
+- Do not delegate any review work directly or indirectly.
+- Do not invoke another review skill or review workflow.
+- Do not use a general-purpose execution unit for review or relabel delegated work as a finder, verifier, audit, compatibility review, verification, or gap sweep.
+- Direct reading, searching, editing, and testing by the current Main Agent remain allowed.
+- `review-report.md` is the only persisted review artifact; do not append separate Review Notes to `context-snapshot.md`.
+
+Missing required evidence, an unsafe repair, a failed verification, or failure to converge by round ten produces `NEEDS_FIXES`.
 
 **Gate Check — Final Review Result**:
 
@@ -16,8 +24,7 @@ Handle the marker:
 
 - `REVIEW_PASS` → proceed to next phase.
 - `REVIEW_NEEDS_FIXES` → log unresolved findings and follow the task failure policy; do not externally restart the review loop.
-- `REVIEW_BLOCKED` → log infrastructure evidence and stop so recovery can retry.
-- `REVIEW_INCOMPLETE` or `REVIEW_INVALID` → inspect the append-only progress once, write `failure-log.md`, and stop. Do not treat an old or incidental verdict as completion and do not synthesize a passing result through Main-Agent self-review.
+- `REVIEW_INCOMPLETE` or `REVIEW_INVALID` → inspect the append-only progress once, write `failure-log.md`, and stop. Do not treat an old or incidental verdict as completion and do not synthesize a passing result outside the skill.
 
 **CP-3**: Review has a valid Final Result and `review-report.md` preserves this execution's progress.