feat(contributor-growth): recent-activity check and stricter pre-filter options for candidate-screen

plugins/magpie-contributor-growth/skills/candidate-screen/SKILL.md

@@ -23,7 +23,7 @@ argument-hint: "[target:committer|pmc|both] [window:6m] [end:YYYY-MM-DD]"
 capability: capability:stats
 surface_hash: sha256:a1767d466ee963d1
 license: Apache-2.0
-measured_tokens: 3723
+measured_tokens: 3994
 ---
 
 <!-- SPDX-License-Identifier: Apache-2.0
@@ -130,7 +130,7 @@ Local modifications go in the override file; framework changes go via PR to `apa
 | `window:Nm` | the configured window, else `6m` | Activity window |
 | `end:YYYY-MM-DD` | today | Last day of the window |
 
-From `<project-config>/contributor-nomination-config.md`: `report_repo` (required), `report_path` (default `reports/`), `screen_prefilter_ratio` (default `0.5`).
+From `<project-config>/contributor-nomination-config.md`: `report_repo` (required), `report_path` (default `reports/`), `screen_prefilter_ratio` (default `0.5`), `screen_prefilter_require` (`any` or `all`, default `any`), `screen_prefilter_targets` (`committer` or `both`, default `committer`), `screen_recent_activity_months` (default `4`; `0` turns the check off).
 A `shortlist_max_missing` left over from an earlier version is ignored; say so once at the start of the run.
 Floors come from `<project-config>/committer-readiness.md`, else `contributor-nomination-config.md`, resolved as `contributor-to-committer` Step 1 does.
 Both files are personal configuration, read from the personal layer first and from `.apache-magpie-overrides/` only as a fallback; [they belong in the personal layer](../../../../docs/contributor-growth/README.md#why-the-configuration-is-personal).
@@ -172,13 +172,20 @@ If even a one-day slice exceeds the cap, stop and say so rather than build a par
 
 ## Step 2 — Pre-filter
 
-**`<governance-body>` target:** no pre-filter — the pool is the current committers who are not members, small enough to measure in full.
+The pre-filter only drops the long tail; it is the one place the floors are used.
 
-**Committer target:** for each person in the pool, run two count-only queries — changes landed (`list_authored(person, state: landed, count_only)`) and changes reviewed (`list_reviews_given(person, count_only)`), in the window — reading `total` only.
-Keep the person when either count is at least `screen_prefilter_ratio` × its floor.
-A floor of `0` (an evidence-only metric) is ignored here; it never keeps anyone by itself.
-Only these two counts are cheap enough to pre-filter a large pool; list, triage and community activity are measured in Step 3 for everyone who stays.
-Log everyone dropped, with both counts, in `dropped`; nobody leaves the pool silently.
+**Recent activity (both targets).**
+When `screen_recent_activity_months` is above `0`, run one count-only query per person — changes landed in the last that many months of the window (`list_authored(person, state: landed, since: <end> − N months, count_only)`) — and drop anyone whose `total` is `0`.
+Someone who stopped contributing months ago has nothing current for the `<governance-body>` to look at, however busy the start of the window was.
+
+**Counts.**
+For each person still in the pool, run two count-only queries — changes landed (`list_authored(person, state: landed, count_only)`) and changes reviewed (`list_reviews_given(person, count_only)`), in the window — reading `total` only, and compare each with `screen_prefilter_ratio` × its floor for the person's target.
+With `screen_prefilter_require: any`, keep the person when either count reaches its share; with `all`, only when both do.
+A floor of `0` (an evidence-only metric) is ignored here: it never keeps anyone by itself, and with `all` it is not required.
+The count check applies to the committer target, and to the `<governance-body>` target too when `screen_prefilter_targets` is `both`; otherwise the `<governance-body>` pool — the current committers who are not members — is small enough to measure in full and only the recent-activity check applies to it.
+
+Only these counts are cheap enough to pre-filter a large pool; list, triage and community activity are measured in Step 3 for everyone who stays.
+Log everyone dropped, with both counts, the landed count for the recent months when it was checked, and which check dropped them, in `dropped`; nobody leaves the pool silently.
 
 ---
 

plugins/magpie-setup/templates/contributor-nomination-config.md

@@ -57,6 +57,9 @@ and there are no meaningful framework defaults.
 | `report_repo` | TODO: `owner/name` of a **private** repository | Where `candidate-screen` commits its report. The skill refuses unless the GitHub API reports the repository as private; restrict it to `<governance-body>` members. |
 | `report_path` | `reports/` | Directory inside `report_repo` for the reports. |
 | `screen_prefilter_ratio` | `0.5` | `candidate-screen` keeps a contributor for full measurement when merged PRs or reviews reach this share of the floor. |
+| `screen_prefilter_require` | `any` | `any`: one of merged PRs and reviews reaching its share of the floor keeps a contributor; `all`: both must. |
+| `screen_prefilter_targets` | `committer` | `committer`: the count pre-filter applies to the committer pool only; `both`: to the <governance-body> pool as well. |
+| `screen_recent_activity_months` | `4` | Drop anyone with no change landed in the last N months of the window, for both targets; `0` turns the check off. |
 
 ---
 

tools/skill-evals/README.md

@@ -91,7 +91,7 @@ Suites are currently implemented for:
 - **setup-privacy-llm** — 6 cases across 2 suites (step-1-resolve, step-4-gate)
 - **preflight-reconciliation** — 7 cases across 1 suite (step-reconciliation)
 - **contributor-calibrate** — 11 cases across 3 suites (step-1-find-nominations, step-4-propose-floors, step-6-write-configuration)
-- **contributor-candidate-screen** — 14 cases across 6 suites (step-0-gates, step-2-pre-filter, step-3-measure, step-4-write-report, step-5-deliver)
+- **contributor-candidate-screen** — 15 cases across 6 suites (step-0-gates, step-2-pre-filter, step-3-measure, step-4-write-report, step-5-deliver)
 - **pr-management-stack-review** — 11 cases across 2 suites (step-3-structural-findings, step-4-reading-plan); the gate, mechanical findings, verdict and posting are pytest-tested in `tools/pr-management`
 
 ## Prerequisites

tools/skill-evals/evals/contributor-candidate-screen/README.md

@@ -11,7 +11,7 @@ skill. Tests six decision points:
 |---|---|---|
 | `step-0-gates` | A public report repository is refused with no alternative offered; a private one proceeds after the audience is confirmed | 2 |
 | `step-1-pool` | Roster ids mapped to GitHub handles (unmapped ids listed, current committers excluded); a merged-PR search over 1000 results sliced by month, never truncated | 2 |
-| `step-2-pre-filter` | Pre-filter ratio applied; a zero (evidence-only) floor ignored; the PMC pool not pre-filtered; every dropped contributor logged with counts | 3 |
+| `step-2-pre-filter` | Pre-filter ratio applied; a zero (evidence-only) floor ignored; the PMC pool not count-filtered by default; no recent landed change drops a contributor; `require: all` and `targets: both`; every dropped contributor logged with counts | 4 |
 | `step-3-measure` | Everyone who passed the pre-filter goes into the report, alphabetically, with no comparison against the floors | 1 |
 | `step-4-write-report` | Surfacing note at the top; verified real name; no `@`-mentions anywhere in the report; capped counts shown as minimums; people listed alphabetically, never by activity, with no floors-met count or threshold column | 3 |
 | `step-5-deliver` | No write without confirmation; privacy re-checked before the write; nothing posted elsewhere | 3 |

tools/skill-evals/evals/contributor-candidate-screen/step-2-pre-filter/fixtures/case-1-dropped-logged/report.md

@@ -2,6 +2,7 @@
      https://www.apache.org/licenses/LICENSE-2.0 -->
 
 screen_prefilter_ratio: 0.5
+screen_recent_activity_months: 0
 Floors (committer): prs_merged 40, reviews_total 20.
 
 Count-only results for the window:

tools/skill-evals/evals/contributor-candidate-screen/step-2-pre-filter/fixtures/case-2-zero-floor-ignored/report.md

@@ -3,6 +3,7 @@
 
 Target: committer.
 screen_prefilter_ratio: 0.5
+screen_recent_activity_months: 0
 Floors (committer): prs_merged 40, reviews_total 0 (evidence only).
 
 Count-only results for the window:

tools/skill-evals/evals/contributor-candidate-screen/step-2-pre-filter/fixtures/case-3-pmc-pool-not-prefiltered/report.md

@@ -3,6 +3,7 @@
 
 Target: pmc (the pool is the 14 current committers who are not PMC members).
 screen_prefilter_ratio: 0.5
+screen_recent_activity_months: 0
 Floors (pmc): prs_merged 40, reviews_total 100.
 
 Count-only results for the window:

tools/skill-evals/evals/contributor-candidate-screen/step-2-pre-filter/fixtures/case-4-recent-activity-and-all/expected.json

@@ -0,0 +1,23 @@
+{
+  "kept": [
+    "india",
+    "lima"
+  ],
+  "dropped": [
+    {
+      "handle": "juliet",
+      "merged_prs": 60,
+      "reviewed_prs": 40
+    },
+    {
+      "handle": "kilo",
+      "merged_prs": 45,
+      "reviewed_prs": 10
+    },
+    {
+      "handle": "mike",
+      "merged_prs": 33,
+      "reviewed_prs": 175
+    }
+  ]
+}

tools/skill-evals/evals/contributor-candidate-screen/step-2-pre-filter/fixtures/case-4-recent-activity-and-all/report.md

@@ -0,0 +1,18 @@
+<!-- SPDX-License-Identifier: Apache-2.0
+     https://www.apache.org/licenses/LICENSE-2.0 -->
+
+Target: both. Window 2026-04-10 → 2026-10-10.
+screen_prefilter_ratio: 1.0
+screen_prefilter_require: all
+screen_prefilter_targets: both
+screen_recent_activity_months: 4
+Floors (committer): prs_merged 40, reviews_total 20. Floors (pmc): prs_merged 40, reviews_total 100.
+
+Count-only results for the window (committer pool):
+- india: merged PRs 50, reviewed PRs 30, landed since 2026-06-10: 12
+- juliet: merged PRs 60, reviewed PRs 40, landed since 2026-06-10: 0
+- kilo: merged PRs 45, reviewed PRs 10, landed since 2026-06-10: 9
+
+Count-only results for the window (pmc pool — committers not on the PMC):
+- lima: merged PRs 80, reviewed PRs 220, landed since 2026-06-10: 20
+- mike: merged PRs 33, reviewed PRs 175, landed since 2026-06-10: 5

tools/spec-loop/specs/contributor-growth.md

@@ -111,7 +111,10 @@ Adopter config scaffolds live in `plugins/magpie-setup/templates/`
   `tools/skill-evals/evals/contributor-calibrate/`.
 - Skill: `contributor-candidate-screen` — cuts the long tail of recent
   contributors with a deterministic pre-filter against the floors (every
-  drop logged; the only place the floors are used), measures everyone
+  drop logged; the only place the floors are used; configurable as `any`
+  or `all` counts, for the committer pool or both pools, plus a
+  recent-activity check that drops anyone with nothing landed in the
+  last `screen_recent_activity_months`), measures everyone
   who remains — deliberately more than the PMC would consider,
   alphabetically by handle, never scored, ranked, shortlisted or judged —
   and writes a report (the list linked to details, plus a