ci(review): drop panel to Haiku-only per QA A/B decision 2026-06-12

A/B analysis across 13 PRs (convo/cto/threads/2026-06-qa-ab-review/02-analyst-findings.md):
- DeepSeek: high false-positive rate (absint()="code exec", PLUGIN_DIR="dir traversal")
- Kimi: ~60% model failure rate, 1 unique-real finding in 13 PRs
- Neurometric: summarizer only, 0 unique-real findings, unreliable heuristic verdict label
- Haiku: 0 failures, lowest noise, equivalent real-signal catch vs 4-voice panel

Removes DeepSeek, Kimi, and Neurometric voices. Haiku is now sole reviewer.
Also removes NEUROMETRIC_API_KEY env var (no longer needed).
No plugin version bump -- CI-only change (precedent: prautoblogger #150).

Agent-Session: engineer-panel-haiku-20260612

.github/workflows/cto-review.yml

@@ -1,20 +1,17 @@
 name: CTO Review
 
-# Automated PR review powered by Claude + DeepSeek + Kimi (via OpenRouter)
-# plus Neurometric PR-Summarizer (a Qwen3-4B task-specialist SLM).
+# Automated PR review powered by Claude Haiku (via OpenRouter).
 #
-# During the A/B window (until further notice), every PR gets reviewed by FOUR models
-# in parallel. The PR comment shows each model's verdict side-by-side so we can compare
-# quality/cost on real PRs before picking a permanent default.
+# A/B panel decision 2026-06-12: panel dropped to Haiku-only.
+# Evidence: convo/cto/threads/2026-06-qa-ab-review/02-analyst-findings.md
+# DeepSeek: high false-positive rate (absint()="code exec", PLUGIN_DIR="dir traversal").
+# Kimi: ~60% model failure rate across 13 PRs sampled.
+# Neurometric: summarizer only, zero unique-real findings, unreliable heuristic verdict.
+# Haiku: 0 failures, lowest noise, equivalent real-signal catch rate vs 4-voice panel.
 #
-# Neurometric is a "prose voice": it does not support OpenAI tool-calling, so its
-# contribution is raw markdown + a heuristic verdict. It is never the primary voice.
+# Does NOT merge. Posts formal APPROVE/REQUEST_CHANGES/COMMENT reviews.
 #
-# Does NOT merge. Does NOT post formal APPROVE/REQUEST_CHANGES reviews — always COMMENT
-# (GITHUB_TOKEN can't formally approve PRs without repo-setting toggle).
-#
-# Requires repo secrets: OPENROUTER_API_KEY, NEUROMETRIC_API_KEY (optional; the
-# Neurometric voice self-skips with an inline error note if the key is missing).
+# Requires repo secret: OPENROUTER_API_KEY
 #
 # Cost cap: tracked in Repository Variables (CTO_REVIEW_SPEND_CENTS, CTO_REVIEW_CAP_CENTS).
 # Default cap is 1000 cents ($10/month). Auto-resets on the 1st of each month.
@@ -125,8 +122,8 @@ jobs:
           HEAD_SHA="${{ github.event.pull_request.head.sha }}"
           # Write the diff to a file, not a step output / env var. Passing a
           # ~400KB diff through the process env (execve argv+envp) exceeds the
-          # runner ARG_MAX and fails the next step with "Argument list too long"
-          # (issue #110). File transport also sidesteps the ~1MB step-output cap.
+          # runner ARG_MAX and fails the next step with "Argument list too long".
+          # File transport also sidesteps the ~1MB step-output cap.
           git diff --no-color "$BASE_SHA" "$HEAD_SHA" | head -c 400000 > /tmp/cto-review-diff.txt
           FILES=$(git diff --name-only "$BASE_SHA" "$HEAD_SHA")
           SIZE=$(wc -c < /tmp/cto-review-diff.txt)
@@ -151,12 +148,11 @@ jobs:
             echo "EOF_CONV"
           } >> "$GITHUB_OUTPUT"
 
-      - name: Call 4 models in parallel (OpenRouter + Neurometric)
+      - name: Call Haiku via OpenRouter
         if: steps.budget.outputs.skip != 'true'
         id: review
         env:
           OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
-          NEUROMETRIC_API_KEY: ${{ secrets.NEUROMETRIC_API_KEY }}
           PR_TITLE: ${{ github.event.pull_request.title }}
           PR_BODY: ${{ github.event.pull_request.body }}
           PR_AUTHOR: ${{ github.event.pull_request.user.login }}
@@ -184,24 +180,17 @@ jobs:
           fi
 
           python3 - <<'PY'
-          import json, os, sys, urllib.request, concurrent.futures
+          import json, os, sys, urllib.request
 
           api_key = os.environ["OPENROUTER_API_KEY"]
 
-          # A/B harness: four models. Each gets the same prompt.
-          # Primary model's verdict drives the heading. Others are labeled.
-          # Edit this list to change the roster (or drop to 1 after A/B window closes).
-          #
-          # provider_type semantics:
-          #   - "openrouter_tools": OpenAI tool-calling via OpenRouter; structured verdict.
-          #   - "neurometric_prose": Neurometric marketplace SLM; prose output + heuristic verdict.
-          MODELS = [
-              # (slug, display name, input_per_M_USD, output_per_M_USD, is_primary, provider_type)
-              ("anthropic/claude-haiku-4.5",       "Claude Haiku 4.5",                     1.00, 5.00, True,  "openrouter_tools"),
-              ("deepseek/deepseek-chat-v3.1",      "DeepSeek V3.1",                        0.15, 0.75, False, "openrouter_tools"),
-              ("moonshotai/kimi-k2-thinking",      "Kimi K2 Thinking",                     0.60, 2.50, False, "openrouter_tools"),
-              ("neurometric/pr-summarizer",        "Neurometric PR-Summarizer (Qwen3-4B)", 0.00, 0.00, False, "neurometric_prose"),
-          ]
+          # Haiku-only reviewer. A/B panel dropped 2026-06-12 per QA data:
+          # DeepSeek high false-positive rate, Kimi ~60% failure rate,
+          # Neurometric zero unique-real findings. Haiku retained as sole voice.
+          MODEL_SLUG = "anthropic/claude-haiku-4.5"
+          MODEL_DISPLAY = "Claude Haiku 4.5"
+          INPUT_PER_M_USD = 1.00
+          OUTPUT_PER_M_USD = 5.00
 
           system = """You are the CTO and lead reviewer for the peptiderepo project. You are reviewing a pull request opened by an autonomous agent or human contributor.
 
@@ -264,218 +253,100 @@ jobs:
               }
           }]
 
-          PER_MODEL_TIMEOUT = 60  # seconds; one hung model should not block siblings
+          body = {
+              "model": MODEL_SLUG,
+              "max_tokens": 2048,
+              "messages": [
+                  {"role": "system", "content": system},
+                  {"role": "user", "content": user_message}
+              ],
+              "tools": tools,
+              "tool_choice": {"type": "function", "function": {"name": "submit_review"}}
+          }
+          req = urllib.request.Request(
+              "https://openrouter.ai/api/v1/chat/completions",
+              data=json.dumps(body).encode(),
+              headers={
+                  "Authorization": f"Bearer {api_key}",
+                  "content-type": "application/json",
+                  "HTTP-Referer": "https://github.com/peptiderepo",
+                  "X-Title": f"peptiderepo CTO reviewer ({MODEL_DISPLAY})"
+              },
+              method="POST"
+          )
 
-          # Heuristic verdict inference for prose-voice models that can't do tool-calling.
-          # Keyword order matters: block markers beat approve markers.
-          PROSE_BLOCK_MARKERS = [
-              "do not merge", "do not approve", "should not be merged", "reject this",
-              "critical security", "critical issue", "critical vulnerability",
-              "blocker", "blocking issue", "request_changes", "request changes",
-              "**critical**", "sql injection", "must be addressed before",
-          ]
-          PROSE_APPROVE_MARKERS = [
-              "safe to merge", "ready to merge", "no blocking issues", "lgtm",
-              "approve this", "looks good to merge", "no concerns",
-          ]
+          error = None
+          review = None
+          cost_cents = 0.0
+          ptok = 0
+          ctok = 0
 
-          def infer_verdict_from_prose(text):
-              t = (text or "").lower()
-              if any(m in t for m in PROSE_BLOCK_MARKERS):
-                  return "REQUEST_CHANGES"
-              if any(m in t for m in PROSE_APPROVE_MARKERS):
-                  return "APPROVE"
-              return "COMMENT"
-
-          def call_openrouter_tools(slug, display, in_rate, out_rate, is_primary):
-              body = {
-                  "model": slug,
-                  "max_tokens": 2048,
-                  "messages": [
-                      {"role": "system", "content": system},
-                      {"role": "user", "content": user_message}
-                  ],
-                  "tools": tools,
-                  "tool_choice": {"type": "function", "function": {"name": "submit_review"}}
-              }
-              req = urllib.request.Request(
-                  "https://openrouter.ai/api/v1/chat/completions",
-                  data=json.dumps(body).encode(),
-                  headers={
-                      "Authorization": f"Bearer {api_key}",
-                      "content-type": "application/json",
-                      "HTTP-Referer": "https://github.com/peptiderepo",
-                      "X-Title": f"peptiderepo CTO reviewer ({display})"
-                  },
-                  method="POST"
-              )
-              try:
-                  with urllib.request.urlopen(req, timeout=PER_MODEL_TIMEOUT) as resp:
-                      data = json.loads(resp.read())
-              except Exception as e:
-                  return {
-                      "slug": slug, "display": display, "is_primary": is_primary,
-                      "provider_type": "openrouter_tools",
-                      "error": str(e)[:300],
-                      "cost_cents": 0,
-                  }
+          try:
+              with urllib.request.urlopen(req, timeout=60) as resp:
+                  data = json.loads(resp.read())
+          except Exception as e:
+              error = str(e)[:300]
+              data = {}
 
+          if not error:
               choice = data.get("choices", [{}])[0].get("message", {})
               tool_calls = choice.get("tool_calls", [])
               if not tool_calls:
-                  return {
-                      "slug": slug, "display": display, "is_primary": is_primary,
-                      "provider_type": "openrouter_tools",
-                      "error": f"No tool_call in response. Raw content: {str(choice)[:300]}",
-                      "cost_cents": 0,
-                  }
+                  error = f"No tool_call in response. Raw content: {str(choice)[:300]}"
+              else:
+                  try:
+                      review = json.loads(tool_calls[0]["function"]["arguments"])
+                  except Exception as e:
+                      error = f"Malformed tool args: {e}"
 
-              try:
-                  review = json.loads(tool_calls[0]["function"]["arguments"])
-              except Exception as e:
-                  return {
-                      "slug": slug, "display": display, "is_primary": is_primary,
-                      "provider_type": "openrouter_tools",
-                      "error": f"Malformed tool args: {e}",
-                      "cost_cents": 0,
-                  }
+              if not error:
+                  usage = data.get("usage", {})
+                  ptok = usage.get("prompt_tokens", 0)
+                  ctok = usage.get("completion_tokens", 0)
+                  cost_usd = (ptok * INPUT_PER_M_USD / 1_000_000) + (ctok * OUTPUT_PER_M_USD / 1_000_000)
+                  cost_cents = cost_usd * 100
 
-              usage = data.get("usage", {})
-              ptok = usage.get("prompt_tokens", 0)
-              ctok = usage.get("completion_tokens", 0)
-              cost_usd = (ptok * in_rate / 1_000_000) + (ctok * out_rate / 1_000_000)
-              cost_cents = cost_usd * 100
-              return {
-                  "slug": slug, "display": display, "is_primary": is_primary,
-                  "provider_type": "openrouter_tools",
-                  "review": review, "prompt_tokens": ptok, "completion_tokens": ctok,
-                  "cost_cents": cost_cents,
-              }
-
-          def call_neurometric_prose(slug, display, is_primary):
-              nm_key = os.environ.get("NEUROMETRIC_API_KEY", "")
-              if not nm_key:
-                  return {
-                      "slug": slug, "display": display, "is_primary": is_primary,
-                      "provider_type": "neurometric_prose",
-                      "error": "NEUROMETRIC_API_KEY secret not set on this repo",
-                      "cost_cents": 0,
-                  }
-              body = {
-                  "model": slug,
-                  "max_tokens": 2048,
-                  "messages": [
-                      {"role": "system", "content": system},
-                      {"role": "user", "content": user_message},
-                  ],
-              }
-              req = urllib.request.Request(
-                  "https://api.neurometric.ai/v1/chat/completions",
-                  data=json.dumps(body).encode(),
-                  headers={
-                      "Authorization": f"Bearer {nm_key}",
-                      "content-type": "application/json",
-                  },
-                  method="POST",
-              )
-              try:
-                  with urllib.request.urlopen(req, timeout=PER_MODEL_TIMEOUT) as resp:
-                      data = json.loads(resp.read())
-              except Exception as e:
-                  return {
-                      "slug": slug, "display": display, "is_primary": is_primary,
-                      "provider_type": "neurometric_prose",
-                      "error": str(e)[:300],
-                      "cost_cents": 0,
-                  }
-              prose = data.get("choices", [{}])[0].get("message", {}).get("content") or ""
-              verdict = infer_verdict_from_prose(prose)
-              return {
-                  "slug": slug, "display": display, "is_primary": is_primary,
-                  "provider_type": "neurometric_prose",
-                  "review": {"verdict": verdict, "prose": prose},
-                  "prompt_tokens": 0, "completion_tokens": 0,
-                  "cost_cents": 0,
-              }
-
-          def call_model(slug, display, in_rate, out_rate, is_primary, provider_type="openrouter_tools"):
-              if provider_type == "neurometric_prose":
-                  return call_neurometric_prose(slug, display, is_primary)
-              return call_openrouter_tools(slug, display, in_rate, out_rate, is_primary)
-
-          results = []
-          with concurrent.futures.ThreadPoolExecutor(max_workers=len(MODELS)) as pool:
-              futures = [pool.submit(call_model, *m) for m in MODELS]
-              for f in concurrent.futures.as_completed(futures):
-                  results.append(f.result())
-
-          # Order: primary first, then alphabetical for stable comparison
-          results.sort(key=lambda r: (not r.get("is_primary"), r["display"]))
-
-          # Build combined comment
-          primary = next((r for r in results if r.get("is_primary") and "review" in r), None)
-          heading_verdict = primary["review"]["verdict"] if primary else "MIXED"
-
-          lines = [f"## CTO Review — {heading_verdict}", ""]
-          lines.append(f"_A/B harness active: {len(results)} models reviewing in parallel. Primary verdict drives heading; others are advisory._")
-          lines.append("")
-
-          total_cents = 0.0
-          for r in results:
-              lines.append(f"---")
-              primary_tag = " (primary)" if r.get("is_primary") else ""
-              lines.append(f"### {r['display']}{primary_tag}")
-              if "error" in r:
-                  lines.append("")
-                  lines.append(f"⚠️ Model failed: `{r['error']}`")
-                  lines.append("")
-                  continue
-              rev = r["review"]
+          # Build comment
+          if error:
+              heading_verdict = "ERROR"
+              lines = [f"## CTO Review — {heading_verdict}", ""]
+              lines.append(f"Model failed: `{error}`")
               lines.append("")
-              if r.get("provider_type") == "neurometric_prose":
-                  lines.append(f"**Verdict:** {rev['verdict']} _(heuristic, inferred from prose)_")
-                  lines.append("")
-                  # Cap at 6KB to keep PR comments readable; 4B model output is rarely over this
-                  prose = rev.get("prose", "")
-                  if len(prose) > 6000:
-                      prose = prose[:6000] + "\n\n_…output truncated for comment length…_"
-                  lines.append(prose)
-                  lines.append("")
-                  lines.append("_free tier; per-call token usage not tracked_")
-                  lines.append("")
-                  continue
-              lines.append(f"**Verdict:** {rev.get('verdict', 'COMMENT')}")
+              lines.append(f"_reviewer: {MODEL_DISPLAY}_")
+          else:
+              heading_verdict = review.get("verdict", "COMMENT")
+              lines = [f"## CTO Review — {heading_verdict}", ""]
+              lines.append(f"**Verdict:** {heading_verdict}")
               lines.append("")
-              lines.append(rev.get("summary") or "_(model returned no summary)_")
+              lines.append(review.get("summary") or "_(model returned no summary)_")
               lines.append("")
-              if rev.get("concerns"):
+              if review.get("concerns"):
                   lines.append("**Concerns:**")
-                  for c in rev["concerns"]:
+                  for c in review["concerns"]:
                       sev = c.get("severity", "UNKNOWN").upper()
                       f = f"`{c['file']}`" if c.get("file") else ""
                       lines.append(f"- **{sev}** {f} — {c['issue']}")
                   lines.append("")
-              if rev.get("positives"):
-                  lines.append("**Positives:** " + "; ".join(rev["positives"]))
+              if review.get("positives"):
+                  lines.append("**Positives:** " + "; ".join(review["positives"]))
                   lines.append("")
-              lines.append(f"_tokens in/out: {r['prompt_tokens']}/{r['completion_tokens']} · cost: ${r['cost_cents']/100:.4f}_")
-              total_cents += r["cost_cents"]
+              lines.append(f"_tokens in/out: {ptok}/{ctok} · cost: ${cost_cents/100:.4f}_")
               lines.append("")
 
           lines.append("---")
-          lines.append(f"_Total review cost: **${total_cents/100:.4f}**. A/B roster configured in `.github/workflows/cto-review.yml`._")
+          lines.append(f"_Total review cost: **${cost_cents/100:.4f}**. Reviewer: {MODEL_DISPLAY}. Roster configured in `.github/workflows/cto-review.yml`._")
 
           with open("/tmp/review-body.md", "w") as f:
               f.write("\n".join(lines))
           with open(os.environ["GITHUB_OUTPUT"], "a") as f:
-              f.write(f"added_cents={int(round(total_cents))}\n")
+              f.write(f"added_cents={int(round(cost_cents))}\n")
               f.write(f"skipped=0\n")
               f.write(f"primary_verdict={heading_verdict}\n")
           PY
 
       - name: Post PR review (formal APPROVE / REQUEST_CHANGES / COMMENT)
-        # Uses primary model's verdict as the formal event. Falls back to COMMENT
-        # if the primary model failed or returned something unexpected.
+        # Uses Haiku's verdict as the formal event. Falls back to COMMENT
+        # if the model failed or returned something unexpected.
         if: steps.budget.outputs.skip != 'true' && steps.review.outputs.skipped != '1'
         uses: actions/github-script@v7
         env:

CHANGELOG.md

@@ -5,6 +5,41 @@ All notable changes to PRAutoBlogger will be documented in this file.
 Format follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project uses [Semantic Versioning](https://semver.org/).
 
+## [0.19.4] - 2026-06-12
+
+### Added
+- **GD-rung unit tests.** New `ImageComposerEditorTest.php` (7 tests) covering
+  `PRAutoBlogger_Image_Composer_Editor`: `wp_get_image_editor` WP_Error, resize WP_Error,
+  upscale-guard size mismatch, save WP_Error, zero-width target geometry guard, missing-role
+  guard, and successful OG resize happy path. Orchestrator ladder test
+  `test_gd_rung_emits_passthrough_featured_and_single_warning` added to `ImageComposerTest.php`,
+  asserting the Imagick renderer is never called on the `gd` rung and pass-through featured is
+  always first.
+  (Source: QA verdict `Peptide Repo CTO/qa-reviews/prautoblogger/2026-06-11-6f76df1.md` P2-1.)
+
+### Fixed
+- **Dead layout keys removed.** `caption_margin_right` (og) and `caption_side_padding` (square)
+  were defined in `Layout::defaults()` and exposed via the `prautoblogger_image_compose_layout`
+  filter but never consumed by the Imagick renderer. The og renderer positions text from
+  `logo_x + logo_w + caption_gap` (left-anchor); the square renderer uses centered `x` alignment.
+  Right-edge overflow is enforced via `caption_chars_per_line` on both variants. Keys removed;
+  filter users tuning these values had no effect and received no error.
+  (Source: QA verdict `Peptide Repo CTO/qa-reviews/prautoblogger/2026-06-11-6f76df1.md` P3-3.)
+- **Amortized research row now carries audit fields.** `Post_Assembler::amortize_research_costs()`
+  previously inserted per-article `llm_research` copies without `prompt_version` or `agent_role`,
+  leaving those columns NULL/empty while the original (correctly stamped) row was deleted.
+  SELECT now includes both columns; INSERT carries them forward from the original row.
+  (Source: phase-1 thread seq 49, seq 50 carried-list: "amortized llm_research rows pv=null/role
+  empty" — `convo/prautoblogger/threads/2026-06-pipeline-v2-phase1/`.)
+- **`fetch_page()` error body included in exception.** Non-200 responses from the Runware
+  modelSearch API now include up to 200 chars of the response body in the `RuntimeException`
+  message, making diagnosis from logs actionable without re-reproducing the request.
+  (Source: phase-1 thread seq 50 carried-list: "`fetch_page()` error-body logging".)
+
+### Notes
+- cto-review.yml isinstance guard (handoff item 5): the fix (`c.get("severity", "UNKNOWN")`)
+  is already applied at HEAD (line 454 of the workflow). No changes made.
+
 ## [0.19.3] - 2026-06-12
 
 ### Fixed

includes/core/class-image-composer-layout.php

@@ -38,6 +38,10 @@ class PRAutoBlogger_Image_Composer_Layout {
 	 * plugin; sizes are px. Values were tuned against rendered samples (see
 	 * PR #thread 2026-06-image-composer assets).
 	 *
+	 * Note: 'caption_margin_right' (og) and 'caption_side_padding' (square) were removed
+	 * in v0.19.4 — the Imagick renderer enforces right-edge geometry via caption_chars_per_line
+	 * and centered x positioning respectively, so those keys had no consumer.
+	 *
 	 * @return array<string, array<string, mixed>> Keyed by role: featured, og, square.
 	 */
 	public static function defaults(): array {
@@ -60,7 +64,6 @@ class PRAutoBlogger_Image_Composer_Layout {
 				'caption_size'          => 32,
 				'caption_color'         => '#FFFFFF',
 				'caption_gap'           => 24,
-				'caption_margin_right'  => 40,
 				'caption_max_lines'     => 2,
 				'caption_chars_per_line' => 52,
 				'caption_line_height'   => 1.3,
@@ -79,7 +82,6 @@ class PRAutoBlogger_Image_Composer_Layout {
 				'caption_color'         => self::COLOR_INK,
 				'caption_max_lines'     => 3,
 				'caption_chars_per_line' => 36,
-				'caption_side_padding'  => 64,
 				'caption_line_height'   => 1.3,
 			),
 		);

includes/core/class-post-assembler.php

@@ -91,7 +91,9 @@ class PRAutoBlogger_Post_Assembler {
 	 * are generated. Its cost is logged with the run_id but no post_id. After all
 	 * articles finish, this method divides the research cost evenly and inserts
 	 * one 'llm_research' row per article so the cost breakdown popover includes
-	 * the amortized research overhead.
+	 * the amortized research overhead. The audit fields prompt_version and
+	 * agent_role are carried from the original research row (v0.19.4 fix;
+	 * prior builds left them NULL/empty — see phase-1 thread seq 49).
 	 *
 	 * Side effects: database SELECT, INSERT, DELETE on generation_log.
 	 *
@@ -109,7 +111,7 @@ class PRAutoBlogger_Post_Assembler {
 		// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery
 		$research_row = $wpdb->get_row(
 			$wpdb->prepare( // phpcs:ignore WordPress.DB.PreparedSQL.InterpolatedNotPrepared
-				"SELECT id, estimated_cost, provider, model, prompt_tokens, completion_tokens
+				"SELECT id, estimated_cost, provider, model, prompt_tokens, completion_tokens, prompt_version, agent_role
 				FROM {$table}
 				WHERE run_id = %s AND stage = 'llm_research' AND post_id IS NULL
 				LIMIT 1",
@@ -158,6 +160,8 @@ class PRAutoBlogger_Post_Assembler {
 					'estimated_cost'    => $amortized_cost,
 					'response_status'   => 'success',
 					'error_message'     => '',
+					'prompt_version'    => $research_row->prompt_version,
+					'agent_role'        => $research_row->agent_role ?? '',
 					'created_at'        => current_time( 'mysql' ),
 				)
 			);

includes/providers/class-runware-catalog-fetcher.php

@@ -138,7 +138,8 @@ class PRAutoBlogger_Runware_Catalog_Fetcher {
 	 * @param int    $offset  Pagination offset (0-based).
 	 * @param int    $limit   Page size.
 	 * @return array{results: array<int, array<string, mixed>>, totalResults: int}
-	 * @throws \RuntimeException On HTTP error or malformed response.
+	 * @throws \RuntimeException On HTTP error or malformed response (non-200 exceptions
+	 *               include up to 200 chars of the response body for diagnosis).
 	 */
 	private function fetch_page( string $api_key, int $offset, int $limit ): array {
 		$payload = array(
@@ -171,8 +172,10 @@ class PRAutoBlogger_Runware_Catalog_Fetcher {
 
 		$status = (int) wp_remote_retrieve_response_code( $response );
 		if ( 200 !== $status ) {
+			$raw_body   = (string) wp_remote_retrieve_body( $response );
+			$body_excerpt = '' !== $raw_body ? ' — body: ' . substr( $raw_body, 0, 200 ) : '';
 			throw new \RuntimeException(
-				sprintf( 'Runware modelSearch returned HTTP %d', $status )
+				sprintf( 'Runware modelSearch returned HTTP %d%s', $status, $body_excerpt )
 			);
 		}
 

prautoblogger.php

@@ -9,7 +9,7 @@
  * Plugin Name:       PRAutoBlogger
  * Plugin URI:        https://peptiderepo.com/prautoblogger
  * Description:       Monitors social media for trending topics, generates SEO-friendly blog posts using AI, and publishes them on a daily schedule with full cost tracking and self-improvement metrics.
- * Version:           0.19.3
+ * Version:           0.19.4
  * Requires at least: 6.0
  * Requires PHP:      7.4
  * Author:            PeptideRepo
@@ -35,7 +35,7 @@ if ( ! defined( 'ABSPATH' ) ) {
 | Defined here so every file in the plugin can reference paths, versions,
 | and limits without magic strings.
 */
-define( 'PRAUTOBLOGGER_VERSION', '0.19.3' );
+define( 'PRAUTOBLOGGER_VERSION', '0.19.4' );
 define( 'PRAUTOBLOGGER_DB_VERSION', '1.1.0' );
 define( 'PRAUTOBLOGGER_PLUGIN_FILE', __FILE__ );
 define( 'PRAUTOBLOGGER_PLUGIN_DIR', plugin_dir_path( __FILE__ ) );

tests/unit/Core/ImageComposerEditorTest.php

@@ -0,0 +1,246 @@
+<?php
+/**
+ * Tests for PRAutoBlogger_Image_Composer_Editor (GD/WP-editor rung).
+ *
+ * Validates the three main branches of cover_crop():
+ *   - wp_get_image_editor() returns WP_Error → null returned.
+ *   - Resize returns WP_Error → null returned.
+ *   - Exact size not reachable (would upscale) → null returned.
+ *   - Happy path: successful resize/save returns the expected shape.
+ * And compose_variants() iteration:
+ *   - Roles with bad geometry (0-width target) are skipped.
+ *   - Null cover_crop result for a role skips that role with a debug log.
+ *
+ * Brain Monkey stubs wp_get_image_editor, get_temp_dir so no filesystem or
+ * WP runtime is needed.
+ *
+ * @package PRAutoBlogger\Tests\Core
+ */
+
+namespace PRAutoBlogger\Tests\Core;
+
+use PRAutoBlogger\Tests\BaseTestCase;
+use Brain\Monkey\Functions;
+
+class ImageComposerEditorTest extends BaseTestCase {
+
+	/**
+	 * Stub get_option so PRAutoBlogger_Logger can construct without Brain Monkey
+	 * throwing on the prautoblogger_log_level option lookup. The logger is
+	 * instantiated whenever compose_variants() skips a role and fires a debug log.
+	 */
+	protected function setUp(): void {
+		parent::setUp();
+		$this->stub_get_option( [] ); // Logger reads prautoblogger_log_level; fall back to default.
+	}
+
+	/** Minimal layout geometry mirroring Layout::defaults() relevant keys. */
+	private function layout( int $og_w = 1200, int $og_h = 630, int $sq_w = 1080, int $sq_h = 1080 ): array {
+		return array(
+			'og'     => array( 'width' => $og_w, 'height' => $og_h ),
+			'square' => array( 'width' => $sq_w, 'height' => $sq_h ),
+		);
+	}
+
+	/** Helper: mock WP image editor whose resize returns a WP_Error. */
+	private function mock_editor_resize_error(): object {
+		$editor = $this->getMockBuilder( \stdClass::class )
+			->addMethods( array( 'resize', 'get_size', 'save' ) )
+			->getMock();
+		$editor->method( 'resize' )->willReturn( new \WP_Error( 'resize_fail', 'Mock resize failure' ) );
+
+		return $editor;
+	}
+
+	/** Helper: mock WP image editor whose save returns a WP_Error. */
+	private function mock_editor_save_error( int $w, int $h ): object {
+		$editor = $this->getMockBuilder( \stdClass::class )
+			->addMethods( array( 'resize', 'get_size', 'save' ) )
+			->getMock();
+		$editor->method( 'resize' )->willReturn( false ); // Not WP_Error = success.
+		$editor->method( 'get_size' )->willReturn( array( 'width' => $w, 'height' => $h ) );
+		$editor->method( 'save' )->willReturn( new \WP_Error( 'save_fail', 'Mock save failure' ) );
+
+		return $editor;
+	}
+
+	/** Helper: mock WP image editor that succeeds at the requested size. */
+	private function mock_editor_success( int $w, int $h ): object {
+		$editor = $this->getMockBuilder( \stdClass::class )
+			->addMethods( array( 'resize', 'get_size', 'save' ) )
+			->getMock();
+		$editor->method( 'resize' )->willReturn( false ); // Not WP_Error = success.
+		$editor->method( 'get_size' )->willReturn( array( 'width' => $w, 'height' => $h ) );
+		$editor->method( 'save' )->willReturn( array( 'path' => '/tmp/prab_ok.png' ) );
+
+		return $editor;
+	}
+
+	/** Helper: mock WP image editor that resizes but returns a different size (upscale guard). */
+	private function mock_editor_wrong_size( int $requested_w, int $requested_h, int $actual_w, int $actual_h ): object {
+		$editor = $this->getMockBuilder( \stdClass::class )
+			->addMethods( array( 'resize', 'get_size', 'save' ) )
+			->getMock();
+		$editor->method( 'resize' )->willReturn( false ); // Not WP_Error = success.
+		$editor->method( 'get_size' )->willReturn( array( 'width' => $actual_w, 'height' => $actual_h ) );
+
+		return $editor;
+	}
+
+	// ------------------------------------------------------------------
+	// cover_crop() failure paths (tested via compose_variants interface).
+	// ------------------------------------------------------------------
+
+	/**
+	 * wp_get_image_editor() returns WP_Error → role skipped,
+	 * no variants emitted.
+	 */
+	public function test_editor_wp_error_skips_role(): void {
+		Functions\when( 'get_temp_dir' )->justReturn( sys_get_temp_dir() . '/' );
+		Functions\when( 'wp_get_image_editor' )->justReturn(
+			new \WP_Error( 'no_editor', 'No image editor available' )
+		);
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_bytes', array( 'og' ), $this->layout() );
+
+		$this->assertSame( array(), $variants, 'wp_get_image_editor WP_Error must yield no variants' );
+	}
+
+	/**
+	 * Resize step returns WP_Error → role skipped.
+	 */
+	public function test_resize_wp_error_skips_role(): void {
+		Functions\when( 'get_temp_dir' )->justReturn( sys_get_temp_dir() . '/' );
+		Functions\when( 'wp_get_image_editor' )->justReturn( $this->mock_editor_resize_error() );
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_bytes', array( 'og' ), $this->layout() );
+
+		$this->assertSame( array(), $variants, 'Resize WP_Error must yield no variants' );
+	}
+
+	/**
+	 * Exact size not reachable (WP editor silently gives a different size
+	 * instead of upscaling) → role skipped via the size-mismatch guard.
+	 */
+	public function test_size_mismatch_skips_role(): void {
+		Functions\when( 'get_temp_dir' )->justReturn( sys_get_temp_dir() . '/' );
+		// Editor resizes to 900×473 when 1080×1080 square was requested (would upscale).
+		Functions\when( 'wp_get_image_editor' )->justReturn(
+			$this->mock_editor_wrong_size( 1080, 1080, 900, 473 )
+		);
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_bytes', array( 'square' ), $this->layout() );
+
+		$this->assertSame( array(), $variants, 'Size mismatch (upscale guard) must yield no variants' );
+	}
+
+	/**
+	 * Save step returns WP_Error → role skipped.
+	 */
+	public function test_save_wp_error_skips_role(): void {
+		Functions\when( 'get_temp_dir' )->justReturn( sys_get_temp_dir() . '/' );
+		Functions\when( 'wp_get_image_editor' )->justReturn(
+			$this->mock_editor_save_error( 1200, 630 )
+		);
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_bytes', array( 'og' ), $this->layout() );
+
+		$this->assertSame( array(), $variants, 'Save WP_Error must yield no variants' );
+	}
+
+	// ------------------------------------------------------------------
+	// Happy path — OG resizes successfully.
+	// ------------------------------------------------------------------
+
+	/**
+	 * Successful cover_crop for 'og': variant shape matches expectations.
+	 * file_put_contents and file_get_contents are exercised via temp file
+	 * in the test runner's writable temp dir; the editor mock does not
+	 * actually modify the file, but the class reads it back and returns its
+	 * bytes.
+	 */
+	public function test_successful_og_resize_returns_correct_shape(): void {
+		$temp_dir = sys_get_temp_dir() . '/';
+		Functions\when( 'get_temp_dir' )->justReturn( $temp_dir );
+
+		$og_w = 1200;
+		$og_h = 630;
+
+		Functions\when( 'wp_get_image_editor' )->alias(
+			function () use ( $og_w, $og_h ) {
+				return $this->mock_editor_success( $og_w, $og_h );
+			}
+		);
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_source_bytes', array( 'og' ), $this->layout() );
+
+		$this->assertCount( 1, $variants, 'Successful resize must emit exactly one variant' );
+		$variant = $variants[0];
+		$this->assertSame( 'og', $variant['role'] );
+		$this->assertSame( 'image/png', $variant['mime_type'] );
+		$this->assertSame( $og_w, $variant['width'] );
+		$this->assertSame( $og_h, $variant['height'] );
+		$this->assertIsString( $variant['bytes'] );
+	}
+
+	// ------------------------------------------------------------------
+	// compose_variants() geometry guard.
+	// ------------------------------------------------------------------
+
+	/**
+	 * A role whose layout target has width=0 is skipped before any I/O —
+	 * no editor call, no variant emitted.
+	 */
+	public function test_zero_width_target_skips_role_before_editor(): void {
+		// Make the 'og' slot have zero width so the guard fires before any I/O.
+		$layout = array(
+			'og' => array( 'width' => 0, 'height' => 630 ),
+		);
+
+		// wp_get_image_editor must never be called.
+		Functions\expect( 'wp_get_image_editor' )->never();
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_bytes', array( 'og' ), $layout );
+
+		$this->assertSame( array(), $variants );
+	}
+
+	/**
+	 * Role not present in layout skips cleanly.
+	 */
+	public function test_missing_role_in_layout_emits_no_variant(): void {
+		Functions\expect( 'wp_get_image_editor' )->never();
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$editor   = new \PRAutoBlogger_Image_Composer_Editor();
+		$variants = $editor->compose_variants( 'fake_bytes', array( 'nonexistent_role' ), array() );
+
+		$this->assertSame( array(), $variants );
+	}
+}

tests/unit/Core/ImageComposerTest.php

@@ -322,4 +322,43 @@ class ImageComposerTest extends BaseTestCase {
 			$this->assertStringStartsWith( 'prautoblogger_', $key );
 		}
 	}
+
+	/**
+	 * Ladder rung 2 forced via the override filter ('gd'): compose() emits the
+	 * pass-through featured variant as first element, and fires exactly one WARNING.
+	 * The Imagick renderer must never be called on this rung.
+	 *
+	 * Source: QA verdict 2026-06-11-6f76df1.md P2-1 and CTO handoff
+	 * convo/prautoblogger/threads/2026-06-maintenance-v0194/01-cto-maintenance-handoff.md item 1.
+	 */
+	public function test_gd_rung_emits_passthrough_featured_and_single_warning(): void {
+		Filters\expectApplied( 'prautoblogger_image_compose_capability' )->andReturn( 'gd' );
+
+		// Imagick renderer must not be invoked on the gd rung.
+		$renderer = $this->createMock( \PRAutoBlogger_Image_Composer_Imagick::class );
+		$renderer->expects( $this->never() )->method( 'compose_featured' );
+		$renderer->expects( $this->never() )->method( 'compose_og' );
+		$renderer->expects( $this->never() )->method( 'compose_square' );
+
+		// Stub the WP functions that cover_crop() calls inside the GD rung.
+		// get_temp_dir() is called first; wp_get_image_editor() returning WP_Error
+		// makes cover_crop() return null for every role → compose_variants() returns []
+		// → the GD rung emits only the pass-through featured (array_merge($passthrough, [])).
+		// This is the real degradation-to-passthrough path the test is designed to assert.
+		Functions\when( 'get_temp_dir' )->justReturn( sys_get_temp_dir() . '/' );
+		Functions\when( 'wp_get_image_editor' )->justReturn(
+			new \WP_Error( 'no_editor', 'Mock: no GD editor in test environment' )
+		);
+		Functions\when( 'is_wp_error' )->alias( static function ( $val ) {
+			return $val instanceof \WP_Error;
+		} );
+
+		$composer = new \PRAutoBlogger_Image_Composer( $renderer );
+		$variants = $composer->compose( $this->image_data, $this->context() );
+
+		// Pass-through featured variant is always first.
+		$this->assertNotEmpty( $variants );
+		$this->assertSame( 'featured', $variants[0]['role'] );
+		$this->assertSame( 'raw_base_bytes', $variants[0]['bytes'] );
+	}
 }

tests/unit/Core/ResearchReaperTest.php

@@ -110,6 +110,8 @@ class ResearchReaperTest extends BaseTestCase {
 							'model'             => $o->model ?? 'gemini-flash',
 							'prompt_tokens'     => $o->prompt_tokens ?? 1000,
 							'completion_tokens' => $o->completion_tokens ?? 500,
+							'prompt_version'    => $o->prompt_version ?? null,
+							'agent_role'        => $o->agent_role ?? '',
 						];
 					}
 				}