prizmkit 1.1.93 → 1.1.95

package/bundled/dev-pipeline/templates/bootstrap-tier2.md

@@ -184,7 +184,7 @@ Wait for Reviewer to return.
 **CP-2**: No CRITICAL issues.
 
 {{IF_CRITIC_ENABLED}}
-### Phase 3.5: Plan Challenge — Critic Agent
+### Phase 3.5: Plan Challenge — Critic Agent(s)
 
 **Guard**: Verify critic agent file exists before spawning:
 ```bash
@@ -192,9 +192,13 @@ ls {{CRITIC_SUBAGENT_PATH}} 2>/dev/null && echo "CRITIC:READY" || echo "CRITIC:M
 ```
 If CRITIC:MISSING — skip Phase 3.5 entirely and proceed to Phase 4. Log: "Critic agent not installed — skipping Plan Challenge."
 
+**Choose ONE path based on `{{CRITIC_COUNT}}`:**
+
+**If {{CRITIC_COUNT}} = 1 → Single Critic** (skip to CP-2.5 after this):
+
 Spawn Critic agent (Agent tool, subagent_type="prizm-dev-team-critic", mode="plan", run_in_background=false).
 
-Spawn failure cap: for team/config/lock errors, retry at most once for this Critic spawn. If the second attempt fails, do not poll for `challenge-report.md`; perform the plan challenge inline and record the fallback.
+Spawn failure cap: for team/config/lock errors, retry at most once for this Critic spawn. If the second attempt fails, do not poll for challenge reports; perform the plan challenge inline and record the fallback.
 
 Prompt:
 > "Read {{CRITIC_SUBAGENT_PATH}}. For feature {{FEATURE_ID}} (slug: {{FEATURE_SLUG}}):
@@ -209,6 +213,49 @@ Wait for Critic to return.
 - Read challenge-report.md. For items marked CRITICAL/HIGH: decide whether to adjust plan.md or document why the plan stands.
 - Max 1 plan revision round.
 
+**If {{CRITIC_COUNT}} = 3 → Multi-Critic Voting** (skip Single Critic above):
+
+**IMPORTANT**: Spawn all 3 Critics in PARALLEL (issue all 3 Agent tool calls in the SAME message turn). Do NOT wait for one before spawning the next.
+
+Spawn Critic-A, Critic-B, Critic-C simultaneously (each with mode="plan", run_in_background=false):
+
+Critic-A prompt:
+> "Read {{CRITIC_SUBAGENT_PATH}}. For feature {{FEATURE_ID}} (slug: {{FEATURE_SLUG}}):
+> **MODE: Plan Challenge — Lens: Architecture & Scalability**
+> Focus on: architectural pattern fit, scalability implications, over-engineering risks, component boundary design.
+> 1. Read `.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md` FIRST — Section 3 has TRAPS/RULES, Section 4 has file manifest.
+> 2. Read `.prizmkit/prizm-docs/root.prizm` and relevant L1/L2 docs for affected modules.
+> 3. Read existing source files in the modules this plan touches.
+> 4. Challenge plan.md through your architecture lens.
+> Write `.prizmkit/specs/{{FEATURE_SLUG}}/challenge-report-A.md`."
+
+Critic-B prompt:
+> "Read {{CRITIC_SUBAGENT_PATH}}. For feature {{FEATURE_ID}} (slug: {{FEATURE_SLUG}}):
+> **MODE: Plan Challenge — Lens: Data Model & Edge Cases**
+> Focus on: data model design fit, entity relationships, edge cases in business logic, missing boundary conditions.
+> 1. Read `.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md` FIRST — Section 3 has TRAPS/RULES, Section 4 has file manifest.
+> 2. Read `.prizmkit/prizm-docs/root.prizm` and relevant L1/L2 docs for affected modules.
+> 3. Read existing source files in the modules this plan touches.
+> 4. Challenge plan.md through your data/edge-case lens.
+> Write `.prizmkit/specs/{{FEATURE_SLUG}}/challenge-report-B.md`."
+
+Critic-C prompt:
+> "Read {{CRITIC_SUBAGENT_PATH}}. For feature {{FEATURE_ID}} (slug: {{FEATURE_SLUG}}):
+> **MODE: Plan Challenge — Lens: Security & Performance**
+> Focus on: security attack surface, authentication/authorization gaps, performance bottlenecks, resource leaks.
+> 1. Read `.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md` FIRST — Section 3 has TRAPS/RULES, Section 4 has file manifest.
+> 2. Read `.prizmkit/prizm-docs/root.prizm` and relevant L1/L2 docs for affected modules.
+> 3. Read existing source files in the modules this plan touches.
+> 4. Challenge plan.md through your security/performance lens.
+> Write `.prizmkit/specs/{{FEATURE_SLUG}}/challenge-report-C.md`."
+
+Spawn failure cap: for team/config/lock errors, each Critic retries at most once. If 2 or more Critics fail to spawn after retry, perform the plan challenge inline and record the fallback.
+
+After all critics return, read all 3 reports:
+- Challenge raised by **2/3 or more** critics → **must respond** (adjust plan or justify why not)
+- Challenge raised by **1/3 only** → logged in context-snapshot but not blocking
+- Max 1 plan revision round.
+
 **CP-2.5**: Plan challenges reviewed and resolved.
 {{END_IF_CRITIC_ENABLED}}
 
@@ -254,7 +301,7 @@ Gate requirements:
 - 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.
 
-### Phase 5: Review + Test — Reviewer Subagent
+### Phase 5: Review — Reviewer Subagent
 
 Spawn Reviewer subagent (Agent tool, subagent_type="prizm-dev-team-reviewer", run_in_background=false).
 
@@ -266,8 +313,7 @@ Prompt:
 > 2. Read `.prizmkit/specs/{{FEATURE_SLUG}}/plan.md` for architecture decisions and completed tasks
 > 3. Read `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt`, then read the referenced `/prizmkit-test` report. Review generated/updated tests, `In-Scope Failures`, `Baseline Failures`, `Out of Scope Gaps`, boundary coverage if present, and the report verdict.
 > 4. Run /prizmkit-code-review with artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/. The skill runs an internal review-fix loop (Reviewer → filter → Dev fix, max 3 rounds) and writes review-report.md.
-> 5. Run the full test suite using `{{TEST_CMD}}`. When running: `({{TEST_CMD}}) 2>&1 | tee /tmp/review-test-out.txt | tail -20`, then grep the file for details — do NOT re-run the suite multiple times.
-> 6. review-report.md will be written to .prizmkit/specs/{{FEATURE_SLUG}}/ by prizmkit-code-review.
+> 5. Do NOT run test suites — the Orchestrator already ran prizmkit-test. Focus on code review and fix strategy.
 > Report: verdict (PASS/NEEDS_FIXES), number of rounds, findings fixed/rejected."
 
 Wait for Reviewer to return.
@@ -287,7 +333,7 @@ Read `review-report.md` and check the Verdict:
 - `PASS` → proceed to next phase
 - `NEEDS_FIXES` → the skill exhausted its max rounds; log the remaining findings and proceed
 
-**CP-3**: Tests pass, review complete.
+**CP-3**: Review complete.
 
 {{IF_BROWSER_INTERACTION}}
 ### Phase 5.5: Browser Verification — MANDATORY
@@ -551,7 +597,7 @@ If you encounter an unrecoverable error, context overflow, or are about to exit
 
 ## Reminders
 
-- Tier 2: orchestrator builds context+plan, Analyzer checks consistency, Dev implements, Reviewer reviews+tests — use direct Agent spawn for agents
+- Tier 2: orchestrator builds context+plan, Analyzer checks consistency, Dev implements, Reviewer reviews — use direct Agent spawn for agents
 - Build context-snapshot.md FIRST; all subagents read it instead of re-reading source files
 - context-snapshot.md is append-only: orchestrator writes Sections 1-4, Dev appends Implementation Log, Reviewer appends Review Notes
 - Gate checks enforce Implementation Log and Review Notes are written before proceeding

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

@@ -234,7 +234,9 @@ Prompt:
 
 **If {{CRITIC_COUNT}} = 3 → Multi-Critic Voting** (skip Single Critic above):
 
-Spawn 3 Critic agents sequentially (each with mode="plan", run_in_background=false), each with a different focus lens:
+**IMPORTANT**: Spawn all 3 Critics in PARALLEL (issue all 3 Agent tool calls in the SAME message turn). Do NOT wait for one before spawning the next.
+
+Spawn Critic-A, Critic-B, Critic-C simultaneously (each with mode="plan", run_in_background=false):
 
 Critic-A prompt (append to base prompt above):
 > "**Focus Lens: Architecture & Scalability.** Prioritize: architectural pattern fit, scalability implications, over-engineering risks, component boundary design.
@@ -248,6 +250,8 @@ Critic-C prompt (append to base prompt above):
 > "**Focus Lens: Security & Performance.** Prioritize: security attack surface, authentication/authorization gaps, performance bottlenecks, resource leaks.
 > Write `.prizmkit/specs/{{FEATURE_SLUG}}/challenge-report-C.md`."
 
+Spawn failure cap: for team/config/lock errors, each Critic retries at most once. If 2 or more Critics fail to spawn after retry, perform the plan challenge inline and record the fallback.
+
 After all critics return, read all 3 reports:
 - Challenge raised by **2/3 or more** critics → **must respond** (adjust plan or justify why not)
 - Challenge raised by **1/3 only** → logged in context-snapshot but not blocking
@@ -315,7 +319,7 @@ Gate requirements:
 - 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.
 
-### Phase 5: Review + Test — Reviewer Agent
+### Phase 5: Review — Reviewer Agent
 
 Spawn Reviewer agent (Agent tool, subagent_type="prizm-dev-team-reviewer", run_in_background=false).
 
@@ -327,8 +331,7 @@ Prompt:
 > 2. Read `.prizmkit/specs/{{FEATURE_SLUG}}/plan.md` for architecture decisions and completed tasks
 > 3. Read `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt`, then read the referenced `/prizmkit-test` report. Review generated/updated tests, `In-Scope Failures`, `Baseline Failures`, `Out of Scope Gaps`, boundary coverage if present, and the report verdict.
 > 4. Run /prizmkit-code-review with artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/. The skill runs an internal review-fix loop (Reviewer → filter → Dev fix, max 3 rounds) and writes review-report.md.
-> 5. Run the full test suite using `{{TEST_CMD}}`. When running tests: `({{TEST_CMD}}) 2>&1 | tee /tmp/review-test-out.txt | tail -20`, then grep `/tmp/review-test-out.txt` for details — do NOT re-run the suite multiple times.
-> 6. review-report.md will be written to .prizmkit/specs/{{FEATURE_SLUG}}/ by prizmkit-code-review.
+> 5. Do NOT run test suites — the Orchestrator already ran prizmkit-test. Focus on code review and fix strategy.
 > Report: verdict (PASS/NEEDS_FIXES), number of rounds, findings fixed/rejected."
 
 Wait for Reviewer to return.
@@ -348,7 +351,7 @@ Read `review-report.md` and check the Verdict:
 - `PASS` → proceed to next phase
 - `NEEDS_FIXES` → the skill exhausted its max rounds; log the remaining findings and proceed
 
-**CP-3**: Integration tests pass, review complete.
+**CP-3**: Review complete.
 
 {{IF_BROWSER_INTERACTION}}
 ### Phase 5.5: Browser Verification — MANDATORY
@@ -629,7 +632,7 @@ If you encounter an unrecoverable error, context overflow, or are about to exit
 
 ## Reminders
 
-- Tier 3: full team — Dev (implementation) → Reviewer (review + test) — spawn agents directly via Agent tool
+- Tier 3: full team — Dev (implementation) → Reviewer (review) — spawn agents directly via Agent tool
 - context-snapshot.md is append-only: orchestrator writes Sections 1-4, Dev appends 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 agents

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

@@ -32,7 +32,7 @@ If CRITIC:MISSING — skip this phase entirely and proceed. Log: "Critic agent n
 
 **If {{CRITIC_COUNT}} = 3 → Multi-Critic Voting** (skip Single Critic above):
 
-Spawn 3 Critic agents sequentially (each with mode="plan", run_in_background=false), each with a different focus lens:
+Spawn 3 Critic agents in PARALLEL (issue all 3 Agent tool calls in the SAME message turn). Do NOT wait for one before spawning the next. Each with mode="plan", run_in_background=false:
 
 Critic-A prompt (append to base prompt above):
 > "**Focus Lens: Architecture & Scalability.** Prioritize: architectural pattern fit, scalability implications, over-engineering risks, component boundary design.

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

@@ -23,9 +23,7 @@ Read `review-report.md` and check the Verdict:
 - `PASS` → proceed to next phase
 - `NEEDS_FIXES` → the skill exhausted its max rounds; log the remaining findings and proceed (do not retry externally)
 
-Run the full test suite: `({{TEST_CMD}}) 2>&1 | tee /tmp/review-test-out.txt | tail -20`
-
-**CP-3**: Review complete, report written, and the Test Failure Recovery Protocol's Success Rule is satisfied.
+**CP-3**: Review complete, report written.
 
 
 **Checkpoint update**: Run the update script to set step `prizmkit-code-review` to `"completed"`:

package/bundled/dev-pipeline-windows/assets/prizm-dev-team-integration.md

@@ -18,9 +18,10 @@ dev-pipeline (outer loop)
             ├── Phase 0: Init (Orchestrator)
             ├── Phase 1-2: Context snapshot + Specify + Plan (Orchestrator)
             ├── Phase 3: Analyze (Reviewer agent) [tier2] / Phase 4: Analyze [tier3]
+            ├── Phase 3.5: Plan Challenge (Critic agent(s), 1 or 3 parallel) [tier2+]
             ├── Phase 4: Implement (Dev agent) [tier2] / Phase 5: Implement [tier3]
-            ├── Phase 4.5/6: Review (Reviewer agent)
-            └── Phase 5/7: Retrospective & Commit (Orchestrator)
+            ├── Phase 5: Review (Reviewer agent)
+            └── Phase 6: Retrospective & Commit (Orchestrator)
             │
             └── Write session-status.json → exit
 ```

package/bundled/dev-pipeline-windows/lib/pipeline.ps1

@@ -291,12 +291,12 @@ function Invoke-PrizmPipeline {
   }
 
   function Write-PrizmStaleKillMarker {
-    param([string]$MarkerPath, [int]$StaleSeconds, [int]$Threshold, [string]$Reason = 'stale_session')
+    param([string]$MarkerPath, [int]$StaleSeconds, [int]$Threshold)
     $markerDir = Split-Path $MarkerPath -Parent
     if ($markerDir) { New-Item -ItemType Directory -Force -Path $markerDir | Out-Null }
     [ordered]@{
       killed_at = (Get-Date).ToUniversalTime().ToString('yyyy-MM-ddTHH:mm:ssZ')
-      reason = $Reason
+      reason = 'stale_session'
       stale_seconds = $StaleSeconds
       threshold = $Threshold
     } | ConvertTo-Json -Compress | Set-Content -Path $MarkerPath -Encoding UTF8
@@ -726,11 +726,9 @@ function Invoke-PrizmPipeline {
 
     $elapsedSeconds = 0
     $staleSeconds = 0
-    $staleReason = 'stale_session'
     $previousLogSize = 0
     $previousProgressSignature = ''
     $previousChildActivitySignature = ''
-    $previousErrorLoopSignature = ''
     $staleKillMarker = Join-Path $logsDir 'stale-kill.json'
     $wasStaleKilled = $false
     while ($true) {
@@ -755,42 +753,9 @@ function Invoke-PrizmPipeline {
 
       $effectiveStaleKillThreshold = Get-PrizmEffectiveStaleKillThreshold -ProgressFile $progressJson -BaseThreshold $staleKillThreshold
 
-      # Check for error-loop: agent is actively producing output but results are
-      # all read-offset errors or wasted calls. Only a newly advancing error
-      # signature counts; stale historical errors must not turn later normal output
-      # into a stale kill.
-      $errorLoopDetected = $false
-      $errorLoopSignature = ''
-      if ($effectiveStaleKillThreshold -gt 0 -and $growth -gt 0 -and (Test-Path $progressJson)) {
-        try {
-          $progress = Get-Content $progressJson -Raw | ConvertFrom-Json
-          $errors = $progress.errors
-          if ($errors -is [array] -and $errors.Count -ge 5) {
-            $recent = $errors[-5..-1]
-            $allBad = ($recent | Where-Object {
-              $_.type -in @("read_offset_overflow", "wasted_call")
-            }).Count -eq 5
-            if ($allBad) {
-              $errorLoopSignature = ($recent | ConvertTo-Json -Compress)
-              if ($previousErrorLoopSignature -and $errorLoopSignature -ne $previousErrorLoopSignature) {
-                $errorLoopDetected = $true
-              }
-            }
-          }
-        } catch {
-          # Ignore JSON parse errors — progress file may be incomplete or malformed
-        }
-      }
-      $previousErrorLoopSignature = $errorLoopSignature
-
-      if ($errorLoopDetected) {
-        $staleReason = 'error_loop'
-        $staleSeconds = $effectiveStaleKillThreshold
-      } elseif ($growth -gt 0 -or $childAdvanced -or $progressAdvanced) {
-        $staleReason = 'stale_session'
+      if ($growth -gt 0 -or $childAdvanced -or $progressAdvanced) {
         $staleSeconds = 0
       } else {
-        $staleReason = 'stale_session'
         $staleSeconds += $heartbeatInterval
       }
 
@@ -807,12 +772,8 @@ function Invoke-PrizmPipeline {
       }
       if ($effectiveStaleKillThreshold -gt 0 -and $staleSeconds -ge $effectiveStaleKillThreshold) {
         $wasStaleKilled = $true
-        if ($staleReason -eq 'error_loop') {
-          Write-PrizmWarn "Session stale-killed due to repeated read-offset/wasted-call error loop"
-        } else {
-          Write-PrizmWarn "Session stale-killed (no progress for ${effectiveStaleKillThreshold}s)"
-        }
-        Write-PrizmStaleKillMarker $staleKillMarker $staleSeconds $effectiveStaleKillThreshold $staleReason
+        Write-PrizmWarn "Session stale-killed (no progress for ${effectiveStaleKillThreshold}s)"
+        Write-PrizmStaleKillMarker $staleKillMarker $staleSeconds $effectiveStaleKillThreshold
         Stop-PrizmSessionProcess $pidPath
         if ($staleKillGraceSeconds -gt 0) { Start-Sleep -Seconds $staleKillGraceSeconds }
         break
@@ -835,15 +796,6 @@ function Invoke-PrizmPipeline {
     }
     Stop-PrizmProgressParser $parserProcess
 
-    $staleKillReason = ''
-    if (Test-Path $staleKillMarker) {
-      try {
-        $staleKillData = Get-Content $staleKillMarker -Raw | ConvertFrom-Json
-        if ($staleKillData.PSObject.Properties['reason']) { $staleKillReason = [string]$staleKillData.reason }
-      } catch {
-        $staleKillReason = ''
-      }
-    }
     $wasInfraError = ($exitCode -ne 0 -and (Test-PrizmInfraError -SessionLog $sessionLog -ProgressJson $progressJson))
     $wasAiRuntimeError = Test-PrizmAiRuntimeError -SessionLog $sessionLog -ProgressJson $progressJson
     $semanticCompletion = if ($Kind -eq 'feature' -and $isGitRepository) {
@@ -866,11 +818,7 @@ function Invoke-PrizmPipeline {
       Write-PrizmWarn "AI session failed due to AI CLI/provider infrastructure error"
       Write-PrizmWarn "Infrastructure errors are retried without consuming code retry budget"
     } elseif ($wasStaleKilled -or (Test-Path $staleKillMarker)) {
-      if ($staleKillReason -eq 'error_loop') {
-        Write-PrizmWarn "Session was killed by heartbeat monitor due to repeated read-offset/wasted-call errors"
-      } else {
-        Write-PrizmWarn "Session was stale-killed by heartbeat monitor (no progress for too long)"
-      }
+      Write-PrizmWarn "Session was stale-killed by heartbeat monitor (no progress for too long)"
       Write-PrizmWarn "Heartbeat-killed sessions are treated as failed; dev branch is preserved for inspection"
     } elseif ($exitCode -ne 0) {
       Write-PrizmWarn "AI session exited with code $exitCode"

package/bundled/dev-pipeline-windows/templates/bootstrap-tier2.md

@@ -220,7 +220,7 @@ Wait for Reviewer to return.
 **CP-2**: No CRITICAL issues.
 
 {{IF_CRITIC_ENABLED}}
-### Phase 3.5: Plan Challenge — Critic Agent
+### Phase 3.5: Plan Challenge — Critic Agent(s)
 
 **Guard**: Verify critic agent file exists before spawning:
 ```powershell
@@ -290,7 +290,7 @@ Gate requirements:
 - 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.
 
-### Phase 5: Review + Test — Reviewer Subagent
+### Phase 5: Review — Reviewer Subagent
 
 Spawn Reviewer subagent (Agent tool, subagent_type="prizm-dev-team-reviewer", run_in_background=false).
 
@@ -302,8 +302,7 @@ Prompt:
 > 2. Read `.prizmkit/specs/{{FEATURE_SLUG}}/plan.md` for architecture decisions and completed tasks
 > 3. Read `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt`, then read the referenced `/prizmkit-test` report. Review generated/updated tests, `In-Scope Failures`, `Baseline Failures`, `Out of Scope Gaps`, boundary coverage if present, and the report verdict.
 > 4. Run /prizmkit-code-review with artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/. The skill runs an internal review-fix loop (Reviewer → filter → Dev fix, max 3 rounds) and writes review-report.md.
-> 5. Run the full test suite using `{{TEST_CMD}}`. When running: `& { {{TEST_CMD}} } 2>&1 | Tee-Object (Join-Path $env:TEMP "review-test-out.txt") | Select-Object -Last 20`, then use Select-String on the file for details — do NOT re-run the suite multiple times.
-> 6. review-report.md will be written to .prizmkit/specs/{{FEATURE_SLUG}}/ by prizmkit-code-review.
+> 5. Do NOT run test suites — the Orchestrator already ran prizmkit-test. Focus on code review and fix strategy.
 > Report: verdict (PASS/NEEDS_FIXES), number of rounds, findings fixed/rejected."
 
 Wait for Reviewer to return.
@@ -323,7 +322,7 @@ Read `review-report.md` and check the Verdict:
 - `PASS` → proceed to next phase
 - `NEEDS_FIXES` → the skill exhausted its max rounds; log the remaining findings and proceed
 
-**CP-3**: Tests pass, review complete.
+**CP-3**: Review complete.
 
 {{IF_BROWSER_INTERACTION}}
 ### Phase 5.5: Browser Verification — MANDATORY
@@ -696,7 +695,7 @@ Remove-Item -Force -ErrorAction SilentlyContinue .prizmkit/specs/{{FEATURE_SLUG}
 
 ## Reminders
 
-- Tier 2: orchestrator builds context+plan, Analyzer checks consistency, Dev implements, Reviewer reviews+tests — use direct Agent spawn for agents
+- Tier 2: orchestrator builds context+plan, Analyzer checks consistency, Dev implements, Reviewer reviews — use direct Agent spawn for agents
 - Build context-snapshot.md FIRST; all subagents read it instead of re-reading source files
 - context-snapshot.md is append-only: orchestrator writes Sections 1-4, Dev appends Implementation Log, Reviewer appends Review Notes
 - Gate checks enforce Implementation Log and Review Notes are written before proceeding

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

@@ -266,7 +266,9 @@ Prompt:
 
 **If {{CRITIC_COUNT}} = 3 → Multi-Critic Voting** (skip Single Critic above):
 
-Spawn 3 Critic agents sequentially (each with mode="plan", run_in_background=false), each with a different focus lens:
+**IMPORTANT**: Spawn all 3 Critics in PARALLEL (issue all 3 Agent tool calls in the SAME message turn). Do NOT wait for one before spawning the next.
+
+Spawn Critic-A, Critic-B, Critic-C simultaneously (each with mode="plan", run_in_background=false):
 
 Critic-A prompt (append to base prompt above):
 > "**Focus Lens: Architecture & Scalability.** Prioritize: architectural pattern fit, scalability implications, over-engineering risks, component boundary design.
@@ -280,6 +282,8 @@ Critic-C prompt (append to base prompt above):
 > "**Focus Lens: Security & Performance.** Prioritize: security attack surface, authentication/authorization gaps, performance bottlenecks, resource leaks.
 > Write `.prizmkit/specs/{{FEATURE_SLUG}}/challenge-report-C.md`."
 
+Spawn failure cap: for team/config/lock errors, each Critic retries at most once. If 2 or more Critics fail to spawn after retry, perform the plan challenge inline and record the fallback.
+
 After all critics return, read all 3 reports:
 - Challenge raised by **2/3 or more** critics → **must respond** (adjust plan or justify why not)
 - Challenge raised by **1/3 only** → logged in context-snapshot but not blocking
@@ -351,7 +355,7 @@ Gate requirements:
 - 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.
 
-### Phase 5: Review + Test — Reviewer Agent
+### Phase 5: Review — Reviewer Agent
 
 Spawn Reviewer agent (Agent tool, subagent_type="prizm-dev-team-reviewer", run_in_background=false).
 
@@ -363,8 +367,7 @@ Prompt:
 > 2. Read `.prizmkit/specs/{{FEATURE_SLUG}}/plan.md` for architecture decisions and completed tasks
 > 3. Read `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt`, then read the referenced `/prizmkit-test` report. Review generated/updated tests, `In-Scope Failures`, `Baseline Failures`, `Out of Scope Gaps`, boundary coverage if present, and the report verdict.
 > 4. Run /prizmkit-code-review with artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/. The skill runs an internal review-fix loop (Reviewer → filter → Dev fix, max 3 rounds) and writes review-report.md.
-> 5. Run the full test suite using `{{TEST_CMD}}`. When running tests: `& { {{TEST_CMD}} } 2>&1 | Tee-Object (Join-Path $env:TEMP "review-test-out.txt") | Select-Object -Last 20`, then Select-String `(Join-Path $env:TEMP "review-test-out.txt")` for details — do NOT re-run the suite multiple times.
-> 6. review-report.md will be written to .prizmkit/specs/{{FEATURE_SLUG}}/ by prizmkit-code-review.
+> 5. Do NOT run test suites — the Orchestrator already ran prizmkit-test. Focus on code review and fix strategy.
 > Report: verdict (PASS/NEEDS_FIXES), number of rounds, findings fixed/rejected."
 
 Wait for Reviewer to return.
@@ -384,7 +387,7 @@ Read `review-report.md` and check the Verdict:
 - `PASS` → proceed to next phase
 - `NEEDS_FIXES` → the skill exhausted its max rounds; log the remaining findings and proceed
 
-**CP-3**: Integration tests pass, review complete.
+**CP-3**: Review complete.
 
 {{IF_BROWSER_INTERACTION}}
 ### Phase 5.5: Browser Verification — MANDATORY
@@ -774,7 +777,7 @@ Remove-Item -Force -ErrorAction SilentlyContinue .prizmkit/specs/{{FEATURE_SLUG}
 
 ## Reminders
 
-- Tier 3: full team — Dev (implementation) → Reviewer (review + test) — spawn agents directly via Agent tool
+- Tier 3: full team — Dev (implementation) → Reviewer (review) — spawn agents directly via Agent tool
 - context-snapshot.md is append-only: orchestrator writes Sections 1-4, Dev appends 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 agents

package/bundled/dev-pipeline-windows/templates/sections/phase-critic-plan-full.md

@@ -32,7 +32,7 @@ If CRITIC:MISSING — skip this phase entirely and proceed. Log: "Critic agent n
 
 **If {{CRITIC_COUNT}} = 3 → Multi-Critic Voting** (skip Single Critic above):
 
-Spawn 3 Critic agents sequentially (each with mode="plan", run_in_background=false), each with a different focus lens:
+Spawn 3 Critic agents in PARALLEL (issue all 3 Agent tool calls in the SAME message turn). Do NOT wait for one before spawning the next. Each with mode="plan", run_in_background=false:
 
 Critic-A prompt (append to base prompt above):
 > "**Focus Lens: Architecture & Scalability.** Prioritize: architectural pattern fit, scalability implications, over-engineering risks, component boundary design.

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

@@ -23,9 +23,7 @@ Read `review-report.md` and check the Verdict:
 - `PASS` → proceed to next phase
 - `NEEDS_FIXES` → the skill exhausted its max rounds; log the remaining findings and proceed (do not retry externally)
 
-Run the full test suite: `& { {{TEST_CMD}} } 2>&1 | Tee-Object (Join-Path $env:TEMP "review-test-out.txt") | Select-Object -Last 20`
-
-**CP-3**: Review complete, report written, and the Test Failure Recovery Protocol's Success Rule is satisfied.
+**CP-3**: Review complete, report written.
 
 
 **Checkpoint update**: Run the update script to set step `prizmkit-code-review` to `"completed"`:

package/bundled/rules/_rules-metadata.json

@@ -20,6 +20,10 @@
     "general/cohesive-modeling": {
       "description": "Group co-occurring variables/fields/concepts into a cohesive unit (struct/class/interface)",
       "tags": ["general", "design"]
+    },
+    "general/agent-operational-rules": {
+      "description": "Framework-level operational rules for all agents — Edit failure recovery, Read offset safety, test output hygiene, incremental validation",
+      "tags": ["general", "agent", "operational"]
     }
   },
   "presets": {
@@ -30,14 +34,16 @@
         "prizm/prizm-commit-workflow",
         "prizm/prizm-progressive-loading",
         "general/prefer-linux-commands",
-        "general/cohesive-modeling"
+        "general/cohesive-modeling",
+        "general/agent-operational-rules"
       ]
     },
     "minimal": {
-      "description": "Minimal: progressive loading + general rules only",
+      "description": "Minimal: progressive loading + general rules + operational rules",
       "rules": [
         "prizm/prizm-progressive-loading",
-        "general/prefer-linux-commands"
+        "general/prefer-linux-commands",
+        "general/agent-operational-rules"
       ]
     },
     "none": {

package/bundled/rules/general/agent-operational-rules.md

@@ -0,0 +1,68 @@
+---
+name: agent-operational-rules
+description: Framework-level operational rules for AI agents — handling Edit failures, Read offsets, test output hygiene, and escalation. These apply to all agents regardless of role.
+tags: [general, agent, operational]
+---
+
+# Agent Operational Rules
+
+These rules apply to ALL agents in the PrizmKit ecosystem. They cover common operational patterns that prevent wasted context and infinite loops.
+
+## Edit Failure Recovery
+
+When an Edit call fails with 'String to replace not found':
+1. STOP editing immediately — do NOT retry the same Edit.
+2. Run `grep -n` to locate the exact line of the target text.
+3. Read with offset = max(grep_line - 20, 0), limit = 50.
+4. Copy the exact text from the Read result into the Edit old_string.
+5. Never guess or extrapolate an offset — grep first, then read, then edit.
+
+## Read Offset Safety
+
+Before any Read with offset + limit:
+- Compute offset + limit. If the last tool_result for this file shows it has N lines, offset MUST be < N.
+- Never request an offset >= known file length.
+
+## Stale Line Counts (Large Files)
+
+Before editing a large file (>1000 lines), verify you know its current line count from the most recent tool_result. Old line counts from earlier turns may be stale if you have since edited the file.
+
+## Consecutive Read Failures
+
+If 3 consecutive Read calls to the same file return 'shorter than provided offset' or 'Wasted call':
+- STOP all work on that file.
+- Send ESCALATION with the file path, current line count, and the offsets you attempted.
+- The orchestrator will provide the correct content.
+
+## Test Output Hygiene
+
+**Applies to: Dev Agent and any agent that runs tests.** Does NOT apply to Reviewer, Critic, or other agents that do not execute test commands.
+
+When tests fail:
+- Run the test command provided in your task assignment ONCE (e.g., `$TEST_CMD`, `npm test`, `pytest`, etc.).
+- Capture output with `tee` (macOS/Linux) or `Tee-Object` (Windows/PowerShell) to a temp file.
+- Then grep/Select-String that file for failure details.
+- Never re-run the full test suite just to apply a different filter to its output.
+
+**macOS/Linux**:
+```bash
+($TEST_CMD) 2>&1 | tee /tmp/test-out.txt | tail -20
+grep -n "FAIL\|Error\|assert" /tmp/test-out.txt
+```
+
+**Windows/PowerShell**:
+```powershell
+& { {{TEST_CMD}} } 2>&1 | Tee-Object (Join-Path $env:TEMP "test-out.txt") | Select-Object -Last 20
+Select-String -Path (Join-Path $env:TEMP "test-out.txt") -Pattern "FAIL|Error|assert"
+```
+
+Note: `$TEST_CMD` / `{{TEST_CMD}}` is a placeholder set by the pipeline or orchestrator. Use whatever test command is provided in your task assignment. Do NOT hardcode a specific test tool — projects vary.
+
+## Incremental Validation
+
+**Applies to: Dev Agent.** Does NOT apply to Reviewer, Critic, or read-only agents.
+
+After every 3 successful Edit operations on a file:
+- Run the relevant test command for that file once.
+- Validate your changes compile and behave correctly.
+- Do not defer all testing to the end.

package/bundled/skills/_metadata.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.1.93",
+  "version": "1.1.95",
   "skills": {
     "prizm-kit": {
       "description": "Full-lifecycle dev toolkit. Covers spec-driven development, Prizm context docs, code quality, debugging, deployment, and knowledge management.",

package/bundled/skills/feature-planner/SKILL.md

@@ -266,7 +266,7 @@ For simple incremental planning, skip detailed Phase 2-3 analysis:
 4. Draft features (title + description + acceptance_criteria + dependencies)
 5. Write draft to `.prizmkit/plans/feature-list.draft.json`, then call the generate script:
    ```bash
-   python3 ${SKILL_DIR}/scripts/validate-and-generate.py generate --input .prizmkit/plans/feature-list.draft.json --output .prizmkit/plans/feature-list.json --mode incremental
+   python3 ${SKILL_DIR}/scripts/validate-and-generate.py generate --input .prizmkit/plans/feature-list.draft.json --output .prizmkit/plans/feature-list.json
    ```
 6. If valid → summarize and recommend next step
 7. If invalid → apply fixes to the draft, re-run generate (max 2 attempts, then escalate to full workflow)
@@ -317,7 +317,7 @@ Key requirements:
 1. Write a draft JSON to a temporary path (e.g., `.prizmkit/plans/feature-list.draft.json`)
 2. Call the generate script to validate and produce the final file:
 ```bash
-python3 ${SKILL_DIR}/scripts/validate-and-generate.py generate --input .prizmkit/plans/feature-list.draft.json --output .prizmkit/plans/feature-list.json --mode <new|incremental>
+python3 ${SKILL_DIR}/scripts/validate-and-generate.py generate --input .prizmkit/plans/feature-list.draft.json --output .prizmkit/plans/feature-list.json
 ```
 The script fills in defaults (`$schema`, `created_at`, `created_by`), validates all fields, and writes the final file only if validation passes. If validation fails, fix the draft and retry.
 

package/bundled/skills/feature-planner/references/error-recovery.md

@@ -4,7 +4,7 @@ Structured error handling for validation failures, interrupted sessions, and che
 
 ## Validation Failures
 
-When `python3 scripts/validate-and-generate.py validate --input <file> --mode <mode>` returns errors:
+When `python3 scripts/validate-and-generate.py validate --input <file>` returns errors:
 
 ### Parse validation output
 Script returns JSON with `"valid": false`, `"errors": [...]`, `"warnings": [...]`

package/bundled/skills/feature-planner/references/incremental-feature-planning.md

@@ -83,7 +83,7 @@ Keep dependency correctness as first constraint.
 ### Step 5: Validate
 Run:
 ```bash
-python3 ${SKILL_DIR}/scripts/validate-and-generate.py validate --input feature-list.json --mode incremental
+python3 ${SKILL_DIR}/scripts/validate-and-generate.py validate --input feature-list.json
 ```
 
 Fix and re-run until pass.

package/bundled/skills/feature-planner/references/new-project-planning.md

@@ -55,7 +55,7 @@ Split into `sub_features` when:
 1. Write `feature-list.json`.
 2. Run:
    ```bash
-   python3 ${SKILL_DIR}/scripts/validate-and-generate.py validate --input feature-list.json --mode new
+   python3 ${SKILL_DIR}/scripts/validate-and-generate.py validate --input feature-list.json
    ```
 3. Fix all errors, then re-run.