fix: harden Apec ingestion boundaries

src/job_research/apec/adapter.py

@@ -300,4 +300,7 @@ class ApecAdapter:
                 polling=1000,
                 timeout=15_000,
             )
+            final_url = page.url
+            if not _is_public_apec_detail_url(final_url):
+                raise ValueError(f"ApecAdapter landed on an unexpected URL after redirects: {final_url}")
             return page.content()

src/job_research/apec/normalize.py

@@ -235,6 +235,9 @@ def normalize_apec_listing(
     description_text = _detail_block_text(soup, "Descriptif du poste")
     if description_text is None:
         warnings.append(_warning("description_text", "Description missing from Apec listing"))
+    elif not _has_useful_text(description_text):
+        warnings.append(_warning("description_text", "Description is empty or placeholder text"))
+        description_text = None
 
     if source_job_id is not None:
         normalized_source_job_id = source_job_id
@@ -283,7 +286,7 @@ def normalize_apec_listing(
         company=company_text,
         location=location if _has_useful_text(location) else None,
         contract_type=contract_type if _has_useful_text(contract_type) else None,
-        description_text=description_text or None,
+        description_text=description_text,
         published_at=published_at_value,
         refreshed_at=refreshed_at_value,
         fetched_at=fetched_at,

src/job_research/cli.py

@@ -115,7 +115,11 @@ def build_profile(
     cv_signals = extract_cv_signals(cv_text)
     candidate_profile = build_candidate_profile_output(cv_signals, authored_profile)
 
+    try:
         save_candidate_profile_yaml(out, candidate_profile)
+    except OSError as exc:
+        typer.echo(f"Unable to write candidate profile to {out}: {exc}", err=True)
+        raise typer.Exit(code=1)
 
     typer.echo(f"candidate profile written to {out}")
     warning_count = len(candidate_profile.warnings)

tests/apec/test_adapter.py

@@ -512,6 +512,27 @@ def test_fetch_listing_html_uses_explicit_company_fallback_chain(monkeypatch) ->
     assert ".details-offer-list" in wait_script
 
 
+def test_fetch_listing_html_rejects_redirected_non_apec_urls(monkeypatch) -> None:
+    page = _FakeDetailPage({}, rendered_html="<html>redirected</html>")
+
+    original_goto = page.goto
+
+    def redirecting_goto(url: str, wait_until: str | None = None) -> None:
+        original_goto(url, wait_until=wait_until)
+        page.url = "https://www.apec.fr/candidat/recherche-emploi.html/emploi/recherche-avancee?error=true"
+
+    monkeypatch.setattr(page, "goto", redirecting_goto)
+    monkeypatch.setattr(adapter_module, "_open_public_page", lambda: _fake_open_public_page(page))
+
+    with pytest.raises(ValueError, match="unexpected URL after redirects"):
+        ApecAdapter().fetch_listing_html(
+            "https://www.apec.fr/candidat/recherche-emploi.html/emploi/detail-offre/111"
+        )
+
+    assert len(page.waited_functions) == 1
+    assert page.goto_urls == ["https://www.apec.fr/candidat/recherche-emploi.html/emploi/detail-offre/111"]
+
+
 def test_fetch_listing_html_rejects_non_apec_hosts() -> None:
     adapter = ApecAdapter()
 
@@ -524,12 +545,14 @@ def test_fetch_listing_html_reuses_browser_context_across_calls(monkeypatch) ->
         def __init__(self) -> None:
             self.goto_urls: list[str] = []
             self.default_timeout: int | None = None
+            self.url = ""
 
         def set_default_timeout(self, timeout: int) -> None:
             self.default_timeout = timeout
 
         def goto(self, url: str, wait_until: str | None = None) -> None:
             self.goto_urls.append(url)
+            self.url = url
 
         def wait_for_function(self, function: str, polling: int | None = None, timeout: int | None = None) -> None:
             return None

tests/apec/test_normalize.py

@@ -250,3 +250,43 @@ def test_normalize_apec_listing_records_warning_for_placeholder_company() -> Non
     assert listing.company is None
     assert listing.refreshed_at == "2026-06-02"
     assert [warning.field for warning in listing.warnings] == ["company"]
+
+
+def test_normalize_apec_listing_records_warning_for_placeholder_description_text() -> None:
+    html = """
+    <html>
+      <body>
+        <main class="container-details-offer">
+          <h1>Data Engineer F/H</h1>
+          <div class="card-offer">
+            <div class="ref-offre">Ref. Apec : 178554452W</div>
+            <ul class="details-offer-list mb-20">
+              <li>CLOUD TEMPLE</li>
+              <li>1 <span> CDI </span></li>
+              <li>Puteaux - 92</li>
+            </ul>
+            <p>Publiée le 20/04/2026 Actualisée le 02/06/2026</p>
+          </div>
+          <article class="card card-ents mb-20">
+            <div class="list-hzt mb-20">
+              <span class="ents-name">CLOUD TEMPLE</span>
+            </div>
+          </article>
+          <div class="details-post">
+            <h4>Descriptif du poste</h4>
+            <p>N/A</p>
+          </div>
+        </main>
+      </body>
+    </html>
+    """
+
+    listing = normalize_apec_listing(
+        url="https://example.test/job/123",
+        html=html,
+        fetched_at="2026-06-01T10:00:00Z",
+        source_job_id="job-123",
+    )
+
+    assert listing.description_text is None
+    assert [warning.field for warning in listing.warnings] == ["description_text"]

tests/test_cli.py

@@ -1,5 +1,6 @@
 from subprocess import run
 from textwrap import dedent
+from pathlib import Path
 
 from typer.testing import CliRunner
 
@@ -150,6 +151,78 @@ def test_build_profile_reports_when_no_warnings_are_included(tmp_path) -> None:
     assert "No warnings included." in result.stdout
 
 
+def test_build_profile_reports_output_write_failures_cleanly(tmp_path, monkeypatch) -> None:
+    cv = tmp_path / "cv.txt"
+    cv.write_text(
+        dedent(
+            """
+            Tonio Example
+            Location: France
+            Languages: French, English
+            Skills: Python, SQL
+            Data Engineer at Acme
+            Education: Master of Science at Example University
+            """
+        ).strip(),
+        encoding="utf-8",
+    )
+    profile = tmp_path / "profile.md"
+    profile.write_text(
+        dedent(
+            """
+            # Candidate Profile
+
+            ## Summary
+            Junior data engineer focused on Python and GCP.
+
+            ## Target Roles
+            - Data Engineer
+
+            ## Strengths
+            - Python
+            - SQL
+
+            ## Skills To Emphasize
+            - GCP
+            - BigQuery
+
+            ## Constraints
+            - CDI only
+            - France only
+
+            ## Notes
+            - Slight preference for French listings.
+            """
+        ).strip(),
+        encoding="utf-8",
+    )
+    out = tmp_path / "candidate-profile.yaml"
+
+    original_write_text = Path.write_text
+
+    def flaky_write_text(
+        self: Path,
+        data: str,
+        encoding: str | None = None,
+        errors: str | None = None,
+        newline: str | None = None,
+    ) -> int:
+        if self == out:
+            raise OSError("disk full")
+
+        return original_write_text(self, data, encoding=encoding, errors=errors, newline=newline)
+
+    monkeypatch.setattr(Path, "write_text", flaky_write_text)
+
+    result = CliRunner().invoke(app, ["build-profile", "--cv", str(cv), "--profile", str(profile), "--out", str(out)])
+
+    assert result.exit_code == 1
+    assert not out.exists()
+    assert "Unable to write candidate profile to" in result.stderr
+    assert "disk full" in result.stderr
+    assert "Traceback" not in result.stderr
+
+
 def test_build_profile_rejects_empty_cv_text_before_writing(tmp_path) -> None:
     cv = tmp_path / "cv.txt"
     cv.write_text("   \n", encoding="utf-8")