fix(seo): sanitize malformed link suggestions + clean link report

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/generator/generator.py

@@ -328,16 +328,25 @@ def _bare_ghost_notice(ghost: "GhostPublisher", post: dict) -> str:
     )
 
 
-def _seo_link_summary(seo: dict, n_links: int) -> str:
-    pairs = [(l["phrase"], l["slug"]) for l in seo.get("internal_links", [])]
-    return f"{n_links} insertados" + (f" — {pairs}" if pairs else "")
+def _seo_link_summary(inserted_pairs: list[dict], n_suggestions: int) -> str:
+    """Report ACTUALLY-INSERTED links only ("phrase → /slug/"), with a separate
+    count of well-formed suggestions that were skipped (no verbatim body match).
+    Never surfaces raw rejected suggestions."""
+    n = len(inserted_pairs)
+    skipped = max(0, n_suggestions - n)
+    head = f"{n} insertados" + (f", {skipped} omitidos" if skipped else "")
+    if not inserted_pairs:
+        return head
+    body = "; ".join(f"{p['phrase']} → /{p['slug']}/" for p in inserted_pairs)
+    return f"{head} — {body}"
 
 
-def _seo_live_message(ghost: "GhostPublisher", post: dict, seo: dict, n_links: int) -> str:
+def _seo_live_message(ghost: "GhostPublisher", post: dict, seo: dict, inserted_pairs: list[dict]) -> str:
     """Separate Telegram message for the live autofill path (SEO written to draft)."""
     slug = post.get("slug", "")
     warns = seo.get("seo_warnings") or []
     warn_line = ("\n⚠️ revisar: " + "; ".join(warns)) if warns else ""
+    n_sug = len(seo.get("internal_links", []))
     return (
         f"📤 *Borrador en Ghost — SEO autorrelleno*\n"
         f"Editar: {ghost.url}/ghost/#/editor/post/{post['id']}\n\n"
@@ -346,7 +355,7 @@ def _seo_live_message(ghost: "GhostPublisher", post: dict, seo: dict, n_links: i
         f"• meta desc ({len(seo['meta_description'])}): {seo['meta_description']}\n"
         f"• excerpt: {len(seo['custom_excerpt'])} car.\n"
         f"• tags: {', '.join(seo['tags'])}\n"
-        f"• internal links: {_seo_link_summary(seo, n_links)}"
+        f"• internal links: {_seo_link_summary(inserted_pairs, n_sug)}"
         f"{warn_line}\n\n"
         f"🖼️ *Imagen sugerida:*\n"
         f"`imagefinder --query \"{seo['image_query']}\" --context \"{seo['image_context']}\" --lang {ghost.lang}`\n\n"
@@ -354,11 +363,12 @@ def _seo_live_message(ghost: "GhostPublisher", post: dict, seo: dict, n_links: i
     )
 
 
-def _seo_dryrun_message(ghost: "GhostPublisher", post: dict, seo: dict, n_links: int) -> str:
+def _seo_dryrun_message(ghost: "GhostPublisher", post: dict, seo: dict, inserted_pairs: list[dict]) -> str:
     """Separate Telegram message for DRY-RUN: bare draft created, SEO only proposed."""
     slug = post.get("slug", "")
     warns = seo.get("seo_warnings") or []
     warn_line = ("\n⚠️ revisar: " + "; ".join(warns)) if warns else ""
+    n_sug = len(seo.get("internal_links", []))
     return (
         f"🧪 *DRY-RUN SEO* — draft creado SIN escribir SEO (solo revisión)\n"
         f"Editar: {ghost.url}/ghost/#/editor/post/{post['id']}\n\n"
@@ -367,7 +377,7 @@ def _seo_dryrun_message(ghost: "GhostPublisher", post: dict, seo: dict, n_links:
         f"• meta desc ({len(seo['meta_description'])}): {seo['meta_description']}\n"
         f"• excerpt ({len(seo['custom_excerpt'])} car.): {seo['custom_excerpt']}\n"
         f"• tags: {', '.join(seo['tags'])}\n"
-        f"• internal links que se insertarían: {_seo_link_summary(seo, n_links)}"
+        f"• internal links que se insertarían: {_seo_link_summary(inserted_pairs, n_sug)}"
         f"{warn_line}\n\n"
         f"🖼️ *Imagen sugerida:*\n"
         f"`imagefinder --query \"{seo['image_query']}\" --context \"{seo['image_context']}\" --lang {ghost.lang}`\n\n"
@@ -523,7 +533,7 @@ class OutputGenerator:
                     raise RuntimeError("generate_seo_fields returned None")
                 # Build the SAME html that will be posted, then link it deterministically.
                 html = ghost._build_html(article_md)
-                linked_html, n_links = insert_internal_links(
+                linked_html, inserted_pairs = insert_internal_links(
                     html, seo["internal_links"], menu, lang)
 
                 if mode == "dryrun":
@@ -531,15 +541,15 @@ class OutputGenerator:
                     # surface the proposal so Jose can inspect before trusting writes.
                     result = await ghost.publish_draft(title, full_output)
                     post = result["posts"][0]
-                    self.last_publish_notice = _seo_dryrun_message(ghost, post, seo, n_links)
+                    self.last_publish_notice = _seo_dryrun_message(ghost, post, seo, inserted_pairs)
                 else:
                     result = await ghost.publish_draft(
                         title, full_output, tags=seo["tags"], seo=seo,
                         body_html=linked_html)
                     post = result["posts"][0]
-                    self.last_publish_notice = _seo_live_message(ghost, post, seo, n_links)
+                    self.last_publish_notice = _seo_live_message(ghost, post, seo, inserted_pairs)
                 logger.info("Auto-published blog to Ghost",
-                            mode=mode, post_id=post["id"], links=n_links)
+                            mode=mode, post_id=post["id"], links=len(inserted_pairs))
                 return ""
             except Exception as e:
                 logger.warning("SEO autofill failed; falling back to bare draft",

src/seo/autofill.py

@@ -484,6 +484,7 @@ async def generate_seo_fields(
         base_msgs = [{"role": "user", "content": user}]
         text1 = await _raw(base_msgs)
         fields = _coerce(_parse_json_object(text1), lang)
+        fields["internal_links"] = _sanitize_links(fields["internal_links"], link_menu)
 
         # Validate against the shared engine using the real body (md→html + links).
         body_html = _markdown_to_html(article_text)
@@ -508,6 +509,7 @@ async def generate_seo_fields(
                 # temperature=0 so the shorten instruction binds deterministically.
                 retry = _coerce(_parse_json_object(
                     await _raw(retry_msgs, max_tokens=768, temperature=0.0)), lang)
+                retry["internal_links"] = _sanitize_links(retry["internal_links"], link_menu)
                 rlinked, _ = insert_internal_links(
                     _markdown_to_html(article_text), retry["internal_links"], link_menu, lang)
                 rviol = R.check_post(_synthetic_post(retry, rlinked, title, slug))
@@ -550,6 +552,34 @@ async def generate_seo_fields(
 
 # ─── 3. Deterministic internal-link insertion ──────────────────────────────
 
+def _sanitize_links(suggestions: list[dict], menu: list[dict]) -> list[dict]:
+    """Defense-in-depth: drop malformed link suggestions before insertion/reporting.
+
+    The "phrase" must be SHORT verbatim ARTICLE text — never a slug or a menu
+    title. A suggestion is dropped when its phrase equals its own slug, any menu
+    slug, or any menu title (the slug-as-phrase / title-as-phrase traps). Even if
+    the model ignores the prompt rule, the output stays clean. Drops are logged at
+    debug and never raise; survivors still go through the verbatim-in-body guard.
+    """
+    menu_slugs = {(m.get("slug") or "").strip().casefold() for m in menu}
+    menu_titles = {(m.get("title") or "").strip().casefold() for m in menu}
+    menu_slugs.discard("")
+    menu_titles.discard("")
+    clean: list[dict] = []
+    for s in suggestions:
+        phrase = (s.get("phrase") or "").strip()
+        slug = (s.get("slug") or "").strip()
+        if not phrase or not slug:
+            continue
+        pf = phrase.casefold()
+        if pf == slug.casefold() or pf in menu_slugs or pf in menu_titles:
+            logger.debug("seo.links: dropped malformed suggestion (slug/title as phrase)",
+                         phrase=phrase, slug=slug)
+            continue
+        clean.append({"phrase": phrase, "slug": slug})
+    return clean
+
+
 def insert_internal_links(
     html: str,
     suggestions: list[dict],
@@ -563,10 +593,11 @@ def insert_internal_links(
     (never inside a tag, never inside an existing <a>); word-boundary match (no
     partial words); each phrase/slug used at most once; cap at MAX_INTERNAL_LINKS.
     A phrase that is missing or already linked is silently skipped (never forced,
-    never an error). Returns (modified_html, links_inserted).
+    never an error). Returns (modified_html, inserted_pairs) where inserted_pairs
+    is the list of {phrase, slug} dicts actually wrapped (len == links inserted).
     """
     if not html or not suggestions:
-        return html, 0
+        return html, []
 
     site = SITE_BY_LANG.get(lang, SITE_BY_LANG["en"])
     menu_slugs = {m["slug"] for m in menu if m.get("slug")}
@@ -576,6 +607,7 @@ def insert_internal_links(
     # links and no edits inside tag attributes.
     tokens = re.split(r"(<[^>]+>)", html)
     inserted = 0
+    inserted_pairs: list[dict] = []
     used_phrases: set[str] = set()
     used_slugs: set[str] = set()
 
@@ -612,6 +644,7 @@ def insert_internal_links(
             )
             tokens[i] = tok[:m.start()] + replacement + tok[m.end():]
             inserted += 1
+            inserted_pairs.append({"phrase": phrase, "slug": slug})
             used_phrases.add(phrase)
             used_slugs.add(slug)
             done = True
@@ -620,4 +653,4 @@ def insert_internal_links(
             logger.debug("seo.links: phrase not found / already linked, skipped",
                          phrase=phrase, slug=slug)
 
-    return "".join(tokens), inserted
+    return "".join(tokens), inserted_pairs