feat(engine): flag repos with a degraded PR read in org outputs

iris/analysis/origin_funnel.py

@@ -8,8 +8,10 @@
 Each stage has a conversion rate. Comparing funnels across origins reveals
 where AI-assisted code drops off relative to human code.
 
-Limitation: PR→Merge stage is not measured (Iris only fetches merged PRs).
-Future work could add closed/rejected PR data to fill this gap.
+Limitation: PR→Merge stage is not measured. Iris reads merged, closed and open
+PRs, but the "In PR" stage reads acceptance, which counts only commits in
+merged PRs, and no stage follows PRs that never merge. The closed PRs already
+read could fill this gap.
 """
 
 # AGGREGATOR_OPT_OUT: consumes a finalized ReportMetrics (post-aggregation composition).
@@ -56,8 +58,20 @@ def calculate_origin_funnel(metrics: ReportMetrics) -> FunnelResult | None:
         metrics: ReportMetrics with origin, acceptance, and durability data.
 
     Returns:
-        FunnelResult with per-origin funnels, or None if no origin data.
+        FunnelResult with per-origin funnels, or None if no origin data, or
+        if the PR read was degraded and acceptance is absent: the "In PR"
+        stage reads acceptance, and every later conversion chains from it.
     """
+    # `pr_enrichment_degraded` is the union over PR states, so it alone cannot
+    # say whether the PRs acceptance reads were lost; acceptance itself can.
+    # The aggregator omits it when the merged PRs lost their enrichment, and a
+    # failed list or read leaves no PRs to match commits to: either way "In
+    # PR" would fall back to "every commit is in a PR". That default is kept
+    # only for PRs absent by design (no gh, no GitHub remote), where nothing
+    # is marked degraded.
+    if metrics.pr_enrichment_degraded and not metrics.acceptance_by_origin:
+        return None
+
     origin_dist = metrics.commit_origin_distribution
     if not origin_dist:
         return None

iris/cli.py

@@ -16,6 +16,7 @@
 from iris.ingestion.github_reader import read_pull_requests_with_fallback
 from iris.metrics.aggregator import aggregate
 from iris.models.context import AnalysisContext
+from iris.models.pull_request import DEGRADED_BASIC, DEGRADED_FETCH
 from iris.reports.narrative import generate_narrative
 from iris.reports.writer import write_output
 
@@ -498,10 +499,15 @@ def _tick(label: str) -> None:
     with span("ingestion.pull_requests", {"repo": repo_name}):
         pr_fetch = read_pull_requests_with_fallback(repo, days=args.days)
     prs = pr_fetch.prs
-    if prs:
+    steps = ", ".join(pr_fetch.degraded)
+    if prs and pr_fetch.degraded:
+        print(s["cli_prs_found_degraded"].format(count=len(prs), steps=steps))
+    elif prs:
         print(s["cli_prs_found"].format(count=len(prs)))
+    elif DEGRADED_BASIC in pr_fetch.degraded or DEGRADED_FETCH in pr_fetch.degraded:
+        print(s["cli_prs_failed"].format(steps=steps))
     elif pr_fetch.degraded:
-        print(s["cli_prs_failed"].format(steps=", ".join(pr_fetch.degraded)))
+        print(s["cli_prs_none_degraded"].format(steps=steps))
     else:
         print(s["cli_prs_skipped"])
     _tick("pull requests fetched")
@@ -523,6 +529,7 @@ def _tick(label: str) -> None:
             prs=prs or None,
             external_data=external_data,
             pr_fetch_degraded=pr_fetch.degraded,
+            pr_fetch_degraded_by_state=pr_fetch.degraded_by_state,
         )
     print(s["cli_classified"].format(count=len(commits)))
     _tick("aggregate analysis done")
@@ -553,6 +560,7 @@ def _tick(label: str) -> None:
         recent_metrics = aggregate(
             recent_commits, churn_days=args.churn_days, prs=recent_prs,
             pr_fetch_degraded=pr_fetch.degraded,
+            pr_fetch_degraded_by_state=pr_fetch.degraded_by_state,
         )
         trend = compute_trend_delta(
             baseline=metrics,
@@ -686,10 +694,12 @@ def _tick(label: str) -> None:
         pre_metrics = aggregate(
             pre_commits, churn_days=args.churn_days, prs=pre_prs,
             pr_fetch_degraded=pr_fetch.degraded,
+            pr_fetch_degraded_by_state=pr_fetch.degraded_by_state,
         )
         post_metrics = aggregate(
             post_commits, churn_days=args.churn_days, prs=post_prs,
             pr_fetch_degraded=pr_fetch.degraded,
+            pr_fetch_degraded_by_state=pr_fetch.degraded_by_state,
         )
 
         # Compute day spans from commit date ranges

iris/i18n.py

@@ -35,10 +35,11 @@
     ),
     "pr_enrichment_degraded_caveat": (
         "PR data is incomplete for this run — failed steps: {steps}. Merge "
-        "strategy is reported as unknown when enrichment failed, and flow, "
-        "review and in-PR metrics may be missing or skewed. A transient "
-        "GitHub error usually clears on the next run; if every run fails, "
-        "check that gh is authenticated (gh auth status)."
+        "strategy is reported as unknown when enrichment failed for merged "
+        "PRs, and PR lifecycle (counts, time to merge, cycle time), flow, "
+        "review, in-PR and open-PR metrics may be missing or skewed. A "
+        "transient GitHub error usually clears on the next run; if every run "
+        "fails, check that gh is authenticated (gh auth status)."
     ),
     "system_disclaimer": (
         "Engineering outcomes emerge from system dynamics, domain complexity, "
@@ -338,9 +339,11 @@
 
     # CLI
     "cli_reading_prs": "Fetching pull requests...",
-    "cli_prs_found": "{count} merged PRs found.",
+    "cli_prs_found": "{count} PRs found.",
     "cli_prs_skipped": "skipped (no GitHub remote or gh CLI).",
     "cli_prs_failed": "failed ({steps}) — continuing without PR data.",
+    "cli_prs_found_degraded": "{count} PRs found — PR read incomplete ({steps}).",
+    "cli_prs_none_degraded": "no PRs in the window — PR read incomplete ({steps}).",
     "cli_classifying": "Classifying commits...",
     "cli_classified": "{count} commits classified.",
     "cli_non_code_repo": (
@@ -527,6 +530,10 @@
         "project manifest, so their metrics describe documentation rather than "
         "delivery: {repos}."
     ),
+    "org_pr_read_degraded": (
+        "PR data is incomplete this run for {count} repository(ies) ({repos}): "
+        "their PR counts and PR-derived metrics in this report may be missing or skewed."
+    ),
     "org_ai_impact_title": "AI Impact Across Organization",
     "org_ai_impact_body": (
         "Of {total_repos} repositories, {ai_repos} contain AI-assisted commits. "
@@ -922,10 +929,12 @@
     "pr_enrichment_degraded_caveat": (
         "Os dados de PR estão incompletos nesta execução — passos com falha: "
         "{steps}. A estratégia de merge sai como desconhecida quando o "
-        "enriquecimento falha, e métricas de flow, de review e de commits em "
-        "PR podem faltar ou sair distorcidas. Um erro transitório do GitHub "
-        "costuma sumir na próxima execução; se toda execução falhar, "
-        "verifique se o gh está autenticado (gh auth status)."
+        "enriquecimento dos PRs mergeados falha, e métricas de ciclo de vida "
+        "de PR (contagens, tempo até o merge, cycle time), de flow, de "
+        "review, de commits em PR e de PRs abertos podem faltar ou sair "
+        "distorcidas. Um erro transitório do GitHub costuma sumir na próxima "
+        "execução; se toda execução falhar, verifique se o gh está "
+        "autenticado (gh auth status)."
     ),
     "system_disclaimer": (
         "Resultados de engenharia emergem de dinâmicas sistêmicas, complexidade do domínio "
@@ -1227,9 +1236,11 @@
 
     # CLI
     "cli_reading_prs": "Buscando pull requests...",
-    "cli_prs_found": "{count} PRs merged encontrados.",
+    "cli_prs_found": "{count} PRs encontrados.",
     "cli_prs_skipped": "pulado (sem remote GitHub ou gh CLI).",
     "cli_prs_failed": "falhou ({steps}) — seguindo sem dados de PR.",
+    "cli_prs_found_degraded": "{count} PRs encontrados — leitura de PR incompleta ({steps}).",
+    "cli_prs_none_degraded": "nenhum PR na janela — leitura de PR incompleta ({steps}).",
     "cli_classifying": "Classificando commits...",
     "cli_classified": "{count} commits classificados.",
     "cli_non_code_repo": (
@@ -1417,6 +1428,11 @@
         "de projeto, então suas métricas descrevem documentação e não "
         "entrega: {repos}."
     ),
+    "org_pr_read_degraded": (
+        "Os dados de PR estão incompletos nesta execução para {count} "
+        "repositório(s) ({repos}): as contagens de PR e as métricas derivadas "
+        "de PR neste relatório podem faltar ou sair distorcidas."
+    ),
     "org_ai_impact_title": "Impacto de IA na Organização",
     "org_ai_impact_body": (
         "De {total_repos} repositórios, {ai_repos} contêm commits assistidos por IA. "