session-memory: session-quality filter (WP-0005 T01)

detect/quality.py: is_real_coding_session drops health-checks / smoke-tests / interrupted / trivially-short sessions (event floor, repo present, substantive tool activity, non-trivial prompt). Wired into run_detect so signals only form over real sessions — fixes the abandoned false-positive. [detect.quality] knobs; existing detect/curate fixtures made realistic. 8 new tests; suite 80/80. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
AGENTIC-WP-0001: complete T02 + close bootstrap workplan
2026-06-07 11:07:22 +02:00 · 2026-06-07 10:15:23 +02:00 · 2026-06-07 10:08:52 +02:00 · 2026-06-07 10:00:56 +02:00
16 changed files with 837 additions and 26 deletions
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -11,6 +11,31 @@

 ---

+## Dev Workflow
+
+The deliverable code lives in `session_memory/` (the Helix Forge coding-session
+memory system). It is **pure-stdlib Python 3.11+** — `tomllib`, `sqlite3`,
+`dataclasses`; no third-party runtime dependencies and no build step. `pytest` is
+the only dev dependency. Run everything from the repo root.
+
+| Need | Command |
+|------|---------|
+| Python | `python3` (3.11+ required for `tomllib`; developed on 3.12) |
+| Install deps | none at runtime; for tests: `pip install pytest` (or `uv pip install pytest`) |
+| Test | `python3 -m pytest` (full suite) · `python3 -m pytest tests/test_curate_review.py` (one file) · `-q` for quiet |
+| Lint / build | none configured — keep changes matching surrounding style |
+| Run: ingest sweep | `python3 -m session_memory.ingest` (`--dry-run`, `--config PATH`) |
+| Run: detect | `python3 -m session_memory.detect` (`--json`, `--min-frequency N`) |
+| Run: curate | `python3 -m session_memory.curate` (`--auto-approve`, `--json`) |
+| Config | `session_memory/config.toml`; local store under `session_memory/.store/` (gitignored) |
+
+**Verify a change before declaring it done:** run `python3 -m pytest` (expect all
+green), and for pipeline changes do a live `ingest → detect → curate` pass against
+the local store. See `session_memory/README.md` for the full layout and the
+detect → curate → distribute flow.
+
+---
+
 ## State Hub Integration

 The Custodian State Hub tracks work across all domains. Interact via HTTP REST —
--- a/docs/PRD-helix-forge.md
+++ b/docs/PRD-helix-forge.md
@@ -255,12 +255,26 @@ record:
  three flavors?
 - **OQ3** Where does detection logic run — local batch jobs, hub-side, or a dedicated
  service? What volume do we actually expect?
- **OQ4** Pattern format: how do we keep one agnostic representation while giving each
-  distributor enough to render high-quality native artifacts?
- **OQ5** What's the minimum trustworthy evidence bar before a pattern is allowed to be
-  distributed to live agent environments?
- **OQ6** How do we prevent pattern bloat — too many low-value instructions degrading
-  agent context budgets (cf. the token-budget policy in global instructions)?
+- ~~**OQ4** Pattern format: how do we keep one agnostic representation while giving each
+  distributor enough to render high-quality native artifacts?~~ **Resolved (Phase 2,
+  AGENTIC-WP-0004):** the `SolutionPattern` core is flavor-agnostic (problem,
+  resolutions, scope, provenance) and carries per-flavor knowledge only in a separate
+  `rendering_hints` sub-structure keyed by flavor — distributors read the hints, the
+  core stays neutral. Catalogued as versioned files-first artifacts (FR-U3).
+- ~~**OQ5** What's the minimum trustworthy evidence bar before a pattern is allowed to be
+  distributed to live agent environments?~~ **Resolved (Phase 2):** a two-tier
+  evidence bar (`[curate.gate]`). A *promote* floor (frequency / distinct sessions /
+  cost-impact) admits a candidate as `provisional`; a stricter *distribution* floor
+  (higher frequency, optional cross-flavor requirement, cost-impact) is required to
+  mark a pattern `approved` + `distribution_ready`. Defaults are conservative and
+  config-tunable.
+- ~~**OQ6** How do we prevent pattern bloat — too many low-value instructions degrading
+  agent context budgets (cf. the token-budget policy in global instructions)?~~
+  **Resolved (Phase 2):** a bloat guard flags duplicate (same id) and near-duplicate
+  (same signal-type+locus) candidates at review time, and the catalog dedups
+  structurally on the source-candidate key so re-promotion never multiplies entries.
+  Thin candidates stay `provisional` (not distributed) rather than padding live
+  context.

 ## 13. Risks

--- a/session_memory/README.md
+++ b/session_memory/README.md
@@ -26,7 +26,14 @@ session_memory/
  detect/signals.py    # signal extractors over digests
  detect/cluster.py    # cluster signals -> candidate patterns + cross-flavor flag
  detect/__main__.py   # python -m session_memory.detect (ranked report)
-  config.toml          # store paths, retention caps, sources, repo->domain map
+  curate/schema.py     # SolutionPattern artifact + per-flavor rendering hints
+  curate/catalog.py    # versioned, files-first Pattern Catalog (dedup on id)
+  curate/gating.py     # promotion evidence bar + bloat guard
+  curate/review.py     # discuss/approve/reject -> promote workflow
+  curate/decisions.py  # hub decision audit trail (graceful local-queue fallback)
+  curate/__main__.py   # python -m session_memory.curate (interactive / --auto-approve)
+  catalog/             # the committed Pattern Catalog (source of truth)
+  config.toml          # store paths, retention caps, sources, repo->domain map, curate gate
 ```

 The local store lives under `session_memory/.store/` (gitignored).
@@ -71,6 +78,42 @@ Candidates are persisted to a Tier 2 `patterns` table and are the input to the
 Curate phase (Phase 2). Patterns whose evidence spans more than one agent flavor
 are flagged `[CROSS-FLAVOR]` — the highest-value reuse targets.

+## Curate candidates into the Pattern Catalog
+
+Review detect candidates into versioned **Solution Patterns** held in the
+files-first catalog (`session_memory/catalog/`). The flow is **detect → curate →
+(Phase 3) distribute**; `curate` refreshes candidates by running detect first.
+
+```bash
+python -m session_memory.curate                 # interactive review (a/r/d per candidate)
+python -m session_memory.curate --auto-approve  # batch: promote all that clear the evidence bar
+python -m session_memory.curate --json          # machine-readable result
+```
+
+- **Promotion** writes a `SolutionPattern` file (id = source candidate key, so
+  re-promoting the same candidate dedups; content changes bump the semver and
+  archive the prior version to `<id>.history.jsonl`).
+- The **evidence bar** (`[curate.gate]`) sets two floors: a promote floor and a
+  stricter *distribution* floor. A thin-but-real candidate lands `provisional`;
+  one clearing the distribution floor lands `approved` + `distribution_ready`.
+- A **bloat guard** flags duplicate / near-duplicate candidates so the catalog
+  stays lean.
+- Re-review is **idempotent** — a remembered decision is skipped unless the
+  candidate's evidence changed; a prior reject is not re-surfaced.
+- Each final promote/reject is recorded as a **hub decision**; if the hub is
+  offline the decision is queued to `[curate].decision_queue` for later sync
+  (the same after-the-fact pattern used in Phase 1).
+
+### Curate knobs (`[curate]` / `[curate.gate]` in config.toml)
+
+| Key | Meaning |
+|-----|---------|
+| `catalog_dir` | committed Pattern Catalog dir (source of truth) |
+| `review_log` / `decision_queue` | remembered decisions + pending hub decisions (gitignored) |
+| `min_frequency` / `min_sessions` / `min_cost_impact` | floor to promote at all |
+| `dist_require_cross_flavor` | require cross-flavor evidence to be distribution-eligible |
+| `dist_min_frequency` / `dist_min_cost_impact` | stricter floor for `distribution_ready` |
+
 ## Retention knobs (`[retention]` in config.toml)

 | Key | Meaning |
@@ -86,7 +129,7 @@ exists, except the explicitly-reported hard-cap overflow path.
 ## Tests

 ```bash
-python -m pytest          # 26 tests: schema, adapter, store, digest, retention, ingest
+python -m pytest          # schema, adapters, store, digest, retention, ingest, detect, curate
 ```

 ## Status
@@ -95,5 +138,7 @@ python -m pytest          # 26 tests: schema, adapter, store, digest, retention,
  adapter, ingest sweep.
 - **Phase 1** (AGENTIC-WP-0003): Codex + Grok adapters, multi-file session merge,
  and the Detect pipeline (signals → clustering → cross-flavor candidate patterns).
- **Next — Phase 2 (Curate):** review/approve candidates into a versioned pattern
-  catalog. **Phase 3 (Distribute) / Phase 4 (Measure)** follow per the PRD.
+- **Phase 2** (AGENTIC-WP-0004): Curate — Solution Pattern schema, versioned
+  files-first Pattern Catalog, discuss/approve/reject review with an evidence bar +
+  bloat guard, and hub-decision audit trail.
+- **Next — Phase 3 (Distribute) / Phase 4 (Measure)** follow per the PRD.
--- a/session_memory/catalog/sp-problem-abandoned-outcome.json
+++ b/session_memory/catalog/sp-problem-abandoned-outcome.json
@@ -0,0 +1,79 @@
+{
+  "created_at": "2026-06-07T08:02:03Z",
+  "distribution_ready": true,
+  "id": "sp-problem-abandoned-outcome",
+  "name": "cross-flavor problem: abandoned",
+  "polarity": "problem",
+  "problem": "cross-flavor problem: abandoned",
+  "provenance": {
+    "detected_at": null,
+    "evidence": {
+      "cost_impact": 13.0,
+      "cross_flavor": true,
+      "flavors": [
+        "claude",
+        "grok"
+      ],
+      "frequency": 13,
+      "key": "problem:abandoned:outcome",
+      "locus": "outcome",
+      "polarity": "problem",
+      "repos": [
+        "can-you-assist",
+        "llm-connect"
+      ],
+      "score": 253.5,
+      "sessions": [
+        "claude:0510d5f4-956d-430a-9e89-6abc54f95b6a",
+        "claude:106fd234-949e-470d-a208-fe5ed8f14562",
+        "claude:377aba4f-8bbf-4760-90e9-469486ab0518",
+        "claude:4c606c31-beff-4a41-a325-ef63c9f8fb0e",
+        "claude:5bffe081-39fb-44cd-9966-4006f9235a0e",
+        "claude:60d3c947-eacf-49e9-b12c-ff8eb6b1c20b",
+        "claude:8f50f5b4-fbc4-4abe-9a7c-b25b2a713671",
+        "claude:95b1fe00-5d2e-482f-9618-fddf9cdbeb51",
+        "claude:c3e782ad-96b9-4cf1-9eb5-defdf3578426",
+        "claude:d75b2084-faec-40cf-aaf8-d7e0c026bde6",
+        "claude:f282058a-0a43-4fb8-87fc-1e67eaa3533c",
+        "grok:019e6103-af11-7a92-8e0b-5f40465d8223",
+        "grok:019e611e-0728-77d3-bb7a-8c5983e5058a"
+      ],
+      "signal_type": "abandoned",
+      "title": "cross-flavor problem: abandoned"
+    },
+    "promoted_at": "2026-06-07T08:02:03Z",
+    "source_key": "problem:abandoned:outcome"
+  },
+  "rendering_hints": {
+    "claude": {
+      "note": "TODO: refine rendering",
+      "target": "CLAUDE.md"
+    },
+    "grok": {
+      "note": "TODO: refine rendering",
+      "target": "instructions"
+    }
+  },
+  "resolutions": [
+    {
+      "detail": "",
+      "steps": [],
+      "summary": "TODO: capture the recommended resolution"
+    }
+  ],
+  "schema_version": 1,
+  "scope": {
+    "domains": [],
+    "flavors": [
+      "claude",
+      "grok"
+    ],
+    "repos": [
+      "can-you-assist",
+      "llm-connect"
+    ]
+  },
+  "status": "approved",
+  "updated_at": "2026-06-07T08:02:03Z",
+  "version": "1.0.0"
+}
--- a/session_memory/catalog/sp-problem-budget_overrun-tokens.json
+++ b/session_memory/catalog/sp-problem-budget_overrun-tokens.json
@@ -0,0 +1,78 @@
+{
+  "created_at": "2026-06-07T08:02:03Z",
+  "distribution_ready": true,
+  "id": "sp-problem-budget_overrun-tokens",
+  "name": "problem: budget overrun",
+  "polarity": "problem",
+  "problem": "problem: budget overrun",
+  "provenance": {
+    "detected_at": null,
+    "evidence": {
+      "cost_impact": 27.135,
+      "cross_flavor": false,
+      "flavors": [
+        "claude"
+      ],
+      "frequency": 8,
+      "key": "problem:budget_overrun:tokens",
+      "locus": "tokens",
+      "polarity": "problem",
+      "repos": [
+        "activity-core",
+        "artifact-store",
+        "citation-evidence",
+        "flex-auth",
+        "infospace-bench",
+        "railiance-apps",
+        "vergabe-teilnahme"
+      ],
+      "score": 217.08,
+      "sessions": [
+        "claude:0ef1b45c-5c27-4e20-88b3-37daeaa24eca",
+        "claude:2c0d14e1-d089-4076-bf35-b134737a261d",
+        "claude:6e0d3d68-872b-4d93-bb09-0691e091314b",
+        "claude:8313f946-f008-4e98-9915-31950380e39e",
+        "claude:8fabd5ce-6a20-4412-9a8b-0f0763394a78",
+        "claude:a7b4a9b3-0942-4899-b502-e76b0013fc42",
+        "claude:b4ae9631-a7eb-42a6-acb1-c65b660c4b74",
+        "claude:bbcf1c2b-14be-40e4-826b-4b2b49b9d212"
+      ],
+      "signal_type": "budget_overrun",
+      "title": "problem: budget overrun"
+    },
+    "promoted_at": "2026-06-07T08:02:03Z",
+    "source_key": "problem:budget_overrun:tokens"
+  },
+  "rendering_hints": {
+    "claude": {
+      "note": "TODO: refine rendering",
+      "target": "CLAUDE.md"
+    }
+  },
+  "resolutions": [
+    {
+      "detail": "",
+      "steps": [],
+      "summary": "TODO: capture the recommended resolution"
+    }
+  ],
+  "schema_version": 1,
+  "scope": {
+    "domains": [],
+    "flavors": [
+      "claude"
+    ],
+    "repos": [
+      "activity-core",
+      "artifact-store",
+      "citation-evidence",
+      "flex-auth",
+      "infospace-bench",
+      "railiance-apps",
+      "vergabe-teilnahme"
+    ]
+  },
+  "status": "approved",
+  "updated_at": "2026-06-07T08:02:03Z",
+  "version": "1.0.0"
+}
--- a/session_memory/catalog/sp-success-clean_pass-outcome.json
+++ b/session_memory/catalog/sp-success-clean_pass-outcome.json
@@ -0,0 +1,106 @@
+{
+  "created_at": "2026-06-07T08:02:03Z",
+  "distribution_ready": true,
+  "id": "sp-success-clean_pass-outcome",
+  "name": "cross-flavor success: clean pass",
+  "polarity": "success",
+  "problem": "cross-flavor success: clean pass",
+  "provenance": {
+    "detected_at": null,
+    "evidence": {
+      "cost_impact": 20.0,
+      "cross_flavor": true,
+      "flavors": [
+        "claude",
+        "grok"
+      ],
+      "frequency": 20,
+      "key": "success:clean_pass:outcome",
+      "locus": "outcome",
+      "polarity": "success",
+      "repos": [
+        "activity-core",
+        "agentic-resources",
+        "artifact-store",
+        "can-you-assist",
+        "citation-evidence",
+        "infospace-bench",
+        "issue-facade",
+        "ops-bridge",
+        "railiance-apps",
+        "state-hub",
+        "the-custodian",
+        "vergabe-teilnahme"
+      ],
+      "score": 600.0,
+      "sessions": [
+        "claude:0ef1b45c-5c27-4e20-88b3-37daeaa24eca",
+        "claude:16bdbec4-b018-4902-9fb5-336f8f3d61c8",
+        "claude:2c0d14e1-d089-4076-bf35-b134737a261d",
+        "claude:30dbad62-c042-41f2-80c1-5953a1100e7f",
+        "claude:39dd33b1-d156-4d6a-8c33-c359b6f841d8",
+        "claude:4307eff6-cd39-4189-be58-79a3acb69d6c",
+        "claude:4340b160-2fb6-47d0-897c-3cac0a8855d8",
+        "claude:631de76e-fdee-43b5-b091-7b7675467ad1",
+        "claude:63fd4df2-5add-4748-af21-c1544825e006",
+        "claude:6e0d3d68-872b-4d93-bb09-0691e091314b",
+        "claude:8313f946-f008-4e98-9915-31950380e39e",
+        "claude:8fabd5ce-6a20-4412-9a8b-0f0763394a78",
+        "claude:99e9c5af-043f-4b97-8d92-14189da8716b",
+        "claude:a7b4a9b3-0942-4899-b502-e76b0013fc42",
+        "claude:a9483f07-c9dc-4f71-9fa0-831790ea965e",
+        "claude:b4ae9631-a7eb-42a6-acb1-c65b660c4b74",
+        "claude:eb837dd1-5b8e-472e-b9e1-4537b10e03e6",
+        "claude:ee9e84f2-bc35-4eb5-a7ad-aaec5f31d965",
+        "claude:f1b25697-0e5f-45f0-81d1-af0f1762c438",
+        "grok:019e6122-00c0-79f3-b4e5-9c70b77c015d"
+      ],
+      "signal_type": "clean_pass",
+      "title": "cross-flavor success: clean pass"
+    },
+    "promoted_at": "2026-06-07T08:02:03Z",
+    "source_key": "success:clean_pass:outcome"
+  },
+  "rendering_hints": {
+    "claude": {
+      "note": "TODO: refine rendering",
+      "target": "CLAUDE.md"
+    },
+    "grok": {
+      "note": "TODO: refine rendering",
+      "target": "instructions"
+    }
+  },
+  "resolutions": [
+    {
+      "detail": "",
+      "steps": [],
+      "summary": "TODO: capture the recommended resolution"
+    }
+  ],
+  "schema_version": 1,
+  "scope": {
+    "domains": [],
+    "flavors": [
+      "claude",
+      "grok"
+    ],
+    "repos": [
+      "activity-core",
+      "agentic-resources",
+      "artifact-store",
+      "can-you-assist",
+      "citation-evidence",
+      "infospace-bench",
+      "issue-facade",
+      "ops-bridge",
+      "railiance-apps",
+      "state-hub",
+      "the-custodian",
+      "vergabe-teilnahme"
+    ]
+  },
+  "status": "approved",
+  "updated_at": "2026-06-07T08:02:03Z",
+  "version": "1.0.0"
+}
--- a/session_memory/config.toml
+++ b/session_memory/config.toml
@@ -31,10 +31,19 @@ enabled = true
 root    = "~/.grok/sessions"
 glob    = "*/*/chat_history.jsonl"

+# Detect phase (AGENTIC-WP-0005): quality filter — drop non-coding/trivial sessions
+# before signals form, so health-checks don't mint false-positive patterns.
+[detect.quality]
+min_events      = 20   # below this many events, not a real coding session
+min_substantive = 3    # require >= this many substantive (edit/read/shell) tool calls
+min_prompt_len  = 25   # first prompt shorter than this is treated as trivial
+
 # Curate phase (AGENTIC-WP-0004): catalog location + promotion evidence bar.
 [curate]
-catalog_dir = "session_memory/catalog"               # files-first Pattern Catalog (committed)
-review_log  = "session_memory/.store/reviews.jsonl"  # remembered decisions (gitignored)
+catalog_dir    = "session_memory/catalog"               # files-first Pattern Catalog (committed)
+review_log     = "session_memory/.store/reviews.jsonl"  # remembered decisions (gitignored)
+decision_queue = "session_memory/.store/decisions.queue.jsonl"  # hub decisions pending sync
+state_hub_workstream_id = "b3703684-f60e-42f3-b03e-dabe3e8ce3f4"  # AGENTIC-WP-0004

 # Evidence bar (OQ5): floors to promote at all, and stricter floors to be
 # distribution-eligible (status=approved, distribution_ready=true).
--- a/session_memory/curate/main.py
+++ b/session_memory/curate/main.py
@@ -0,0 +1,130 @@
+"""Curate entrypoint (T06): review detect candidates into the Pattern Catalog.
+
+    python -m session_memory.curate [--config PATH] [--auto-approve] [--json]
+                                    [--workstream-id ID]
+
+Refreshes candidate patterns (runs the detect pipeline), then drives them through
+the review workflow — **interactive** by default, or **batch** with
+``--auto-approve`` (promote everything clearing the evidence bar, reject the rest)
+for kaizen-agent runs. Candidates are presented cross-flavor first (detect's
+ranking). Emits a catalog diff summary and, with ``--json``, a machine-readable
+result. Approvals land in the files-first catalog; each final decision is logged
+as a hub decision (queued if the hub is down).
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+
+from ..detect.__main__ import run_detect
+from ..ingest import _expand, load_config
+from .catalog import Catalog
+from .decisions import DecisionRecorder
+from .gating import bloat_warnings, evaluate, gate_config
+from .review import APPROVE, DISCUSS, REJECT, ReviewLog, review
+
+
+def _curate_paths(config: dict):
+    c = config.get("curate", {})
+    catalog_dir = _expand(c.get("catalog_dir", "session_memory/catalog"))
+    review_log = _expand(c.get("review_log", "session_memory/.store/reviews.jsonl"))
+    queue = _expand(c.get("decision_queue", "session_memory/.store/decisions.queue.jsonl"))
+    ws_id = c.get("state_hub_workstream_id")
+    return catalog_dir, review_log, queue, ws_id
+
+
+def _render_candidate(cand: dict, gate, existing) -> str:
+    g = evaluate(cand, gate)
+    flag = " [CROSS-FLAVOR]" if cand.get("cross_flavor") else ""
+    lines = [
+        f"\n{cand['title']}{flag}",
+        f"  key={cand['key']}  score={cand.get('score')} freq={cand['frequency']} "
+        f"impact={cand.get('cost_impact')}",
+        f"  flavors={','.join(cand.get('flavors', []))}  "
+        f"repos={','.join(cand.get('repos', [])) or '-'}  sessions={len(cand.get('sessions', []))}",
+        f"  gate: promotable={g.promotable} distribution_ready={g.distribution_ready}"
+        + (f"  ({'; '.join(g.reasons)})" if g.reasons else ""),
+    ]
+    for w in bloat_warnings(cand, existing):
+        lines.append(f"  bloat: {w}")
+    return "\n".join(lines)
+
+
+def _interactive_decider(gate, catalog):
+    def decide(cand):
+        print(_render_candidate(cand, gate, catalog.list()))
+        while True:
+            choice = input("  [a]pprove / [r]eject / [d]iscuss ? ").strip().lower()
+            if choice in ("a", "approve"):
+                return (APPROVE, input("  rationale: ").strip() or "approved")
+            if choice in ("r", "reject"):
+                return (REJECT, input("  rationale: ").strip() or "rejected")
+            if choice in ("d", "discuss"):
+                return (DISCUSS, "deferred for discussion")
+    return decide
+
+
+def _auto_decider(gate):
+    """Batch policy: approve candidates clearing the promote floor, reject the rest."""
+    def decide(cand):
+        g = evaluate(cand, gate)
+        if g.promotable:
+            return (APPROVE, "auto-approved: clears evidence bar")
+        return (REJECT, "auto-rejected: " + "; ".join(g.reasons))
+    return decide
+
+
+def _summary(result, n_candidates: int) -> str:
+    added = [k for k, a in result.approved if a in ("added", "versioned", "updated")]
+    lines = [
+        f"# Curate summary  ({n_candidates} candidates reviewed)",
+        f"  approved : {len(result.approved)}  ({', '.join(f'{k}:{a}' for k, a in result.approved) or '-'})",
+        f"  rejected : {len(result.rejected)}  ({', '.join(result.rejected) or '-'})",
+        f"  deferred : {len(result.deferred)}  ({', '.join(result.deferred) or '-'})",
+        f"  skipped  : {len(result.skipped)}  (already decided)",
+        f"  catalog writes: {len(added)}",
+    ]
+    return "\n".join(lines)
+
+
+def main(argv=None) -> int:
+    here = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+    ap = argparse.ArgumentParser(description="Curate detect candidates into the Pattern Catalog.")
+    ap.add_argument("--config", default=os.path.join(here, "config.toml"))
+    ap.add_argument("--auto-approve", action="store_true",
+                    help="batch mode: promote everything clearing the evidence bar")
+    ap.add_argument("--min-frequency", type=int, default=2)
+    ap.add_argument("--workstream-id", default=None, help="hub workstream for decisions")
+    ap.add_argument("--json", action="store_true", help="emit machine-readable JSON")
+    args = ap.parse_args(argv)
+
+    config = load_config(args.config)
+    candidates = run_detect(config, min_frequency=args.min_frequency)
+
+    catalog_dir, review_log_path, queue_path, ws_id = _curate_paths(config)
+    gate = gate_config(config)
+    catalog = Catalog(catalog_dir)
+    log = ReviewLog(review_log_path)
+    recorder = DecisionRecorder(queue_path, workstream_id=args.workstream_id or ws_id)
+
+    decide = _auto_decider(gate) if args.auto_approve else _interactive_decider(gate, catalog)
+    result = review(candidates, decide, catalog, log, gate=gate, recorder=recorder)
+
+    if args.json:
+        print(json.dumps({
+            "approved": result.approved, "rejected": result.rejected,
+            "deferred": result.deferred, "skipped": result.skipped,
+            "decisions_queued": len(recorder.pending()),
+        }, indent=2))
+    else:
+        print(_summary(result, len(candidates)))
+        if recorder.pending():
+            print(f"  decisions queued (hub offline): {len(recorder.pending())} "
+                  f"-> {queue_path}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
--- a/session_memory/detect/main.py
+++ b/session_memory/detect/main.py
@@ -16,13 +16,14 @@ import os
 from ..core.store import Store
 from ..ingest import _expand, load_config
 from .cluster import cluster
+from .quality import filter_real, quality_config
 from .signals import extract_signals


 def run_detect(config: dict, *, min_frequency: int = 2) -> list[dict]:
    store_cfg = config.get("store", {})
    store = Store(_expand(store_cfg["db_path"]), _expand(store_cfg["blob_dir"]))
-    digests = store.list_digests()
+    digests = filter_real(store.list_digests(), quality_config(config))
    signals = extract_signals(digests)
    patterns = [p.to_dict() for p in cluster(signals, min_frequency=min_frequency)]
    store.save_patterns(patterns)
@@ -56,7 +57,8 @@ def main(argv=None) -> int:

    config = load_config(args.config)
    store_cfg = config.get("store", {})
-    n = len(Store(_expand(store_cfg["db_path"]), _expand(store_cfg["blob_dir"])).list_digests())
+    all_digests = Store(_expand(store_cfg["db_path"]), _expand(store_cfg["blob_dir"])).list_digests()
+    n = len(filter_real(all_digests, quality_config(config)))
    patterns = run_detect(config, min_frequency=args.min_frequency)

    if args.json:
--- a/session_memory/detect/quality.py
+++ b/session_memory/detect/quality.py
@@ -0,0 +1,75 @@
+"""Session-quality filter (T01).
+
+The capture layer ingests *every* session it finds — including API health-checks,
+smoke-tests, and interrupted runs (e.g. ``llm-connect`` firing "Say hello in one
+word", or a transcript that is just ``[Request interrupted by user]``). These are
+not real coding work, but the outcome heuristic labels the short ones ``abandoned``
+and the clusterer then mints false-positive "problem" patterns from them.
+
+:func:`is_real_coding_session` gates those out so Detect signals/clusters form only
+over genuine coding sessions. It is intentionally conservative — a session counts
+as real if it shows substantive activity, and is dropped only on clear trivial
+markers. Thresholds come from ``[detect.quality]`` in ``config.toml``.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Optional
+
+# Prompt prefixes/markers that indicate a non-coding or interrupted session.
+_TRIVIAL_PROMPTS = (
+    "say hello", "hello", "[request interrupted", "return only this json",
+    "ping", "ok", "<system-reminder>",
+)
+
+# Tool buckets that count as "substantive" coding activity.
+_SUBSTANTIVE_TOOLS = (
+    "Edit", "Write", "Read", "Bash", "search_replace", "write", "read_file",
+    "run_terminal_command", "grep", "Grep", "glob", "Glob", "NotebookEdit",
+)
+
+
+@dataclass
+class QualityConfig:
+    min_events: int = 20          # below this, not a real coding session
+    min_substantive: int = 3      # >= this many substantive tool calls required
+    min_prompt_len: int = 25      # first prompt shorter than this is suspect
+
+
+def quality_config(config: Optional[dict] = None) -> QualityConfig:
+    d = (config or {}).get("detect", {}).get("quality", {}) if config else {}
+    return QualityConfig(
+        min_events=d.get("min_events", 20),
+        min_substantive=d.get("min_substantive", 3),
+        min_prompt_len=d.get("min_prompt_len", 25),
+    )
+
+
+def _substantive_calls(digest: dict) -> int:
+    hist = digest.get("tool_histogram") or {}
+    return sum(n for t, n in hist.items() if t in _SUBSTANTIVE_TOOLS)
+
+
+def is_real_coding_session(digest: dict, config: Optional[QualityConfig] = None) -> bool:
+    cfg = config or QualityConfig()
+
+    if not digest.get("repo"):
+        return False
+    if digest.get("event_count", 0) < cfg.min_events:
+        return False
+    if _substantive_calls(digest) < cfg.min_substantive:
+        return False
+
+    prompt = (digest.get("first_prompt") or "").strip().lower()
+    if len(prompt) < cfg.min_prompt_len:
+        return False
+    if any(prompt.startswith(p) for p in _TRIVIAL_PROMPTS):
+        return False
+
+    return True
+
+
+def filter_real(digests: list[dict], config: Optional[QualityConfig] = None) -> list[dict]:
+    cfg = config or QualityConfig()
+    return [d for d in digests if is_real_coding_session(d, cfg)]
--- a/tests/test_curate_entrypoint.py
+++ b/tests/test_curate_entrypoint.py
@@ -0,0 +1,84 @@
+"""Curate entrypoint tests (T06): batch auto-approve end-to-end via the store."""
+
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from session_memory.core.store import Store  # noqa: E402
+from session_memory.curate.__main__ import main  # noqa: E402
+from session_memory.curate.catalog import Catalog  # noqa: E402
+
+
+def _digest(uid, flavor, repo, **markers):
+    return {
+        "session_uid": uid, "flavor": flavor, "repo": repo, "outcome": "fail",
+        "cost": {"input_tokens": 10, "output_tokens": 1},
+        "markers": {"errors": markers.get("errors", 0), "retries": markers.get("retries", 0),
+                    "test_runs": 0, "edits": 0, "human_interventions": 0},
+        # real coding session per the quality filter (WP-0005 T01)
+        "event_count": 40, "first_prompt": "Fix the failing build and retry the suite",
+        "tool_histogram": {"Bash": 20, "Edit": 12, "Read": 8},
+    }
+
+
+def _write_config(tmp_path) -> str:
+    store = tmp_path / ".store"
+    catalog = tmp_path / "catalog"
+    cfg = f"""
+[store]
+db_path = "{store / 'm.db'}"
+blob_dir = "{store / 'blobs'}"
+cursor = "{store / 'c.json'}"
+
+[curate]
+catalog_dir = "{catalog}"
+review_log = "{store / 'reviews.jsonl'}"
+decision_queue = "{store / 'decisions.queue.jsonl'}"
+
+[curate.gate]
+min_frequency = 2
+min_sessions = 2
+"""
+    path = tmp_path / "config.toml"
+    path.write_text(cfg)
+    return str(path), str(store), str(catalog)
+
+
+def test_auto_approve_promotes_cross_flavor(tmp_path, capsys):
+    cfg_path, store_dir, catalog_dir = _write_config(tmp_path)
+    st = Store(os.path.join(store_dir, "m.db"), os.path.join(store_dir, "blobs"))
+    st.write_digest("claude:a", _digest("claude:a", "claude", "r1", retries=5))
+    st.write_digest("codex:b", _digest("codex:b", "codex", "r2", retries=4))
+    st.close()
+
+    rc = main(["--config", cfg_path, "--auto-approve"])
+    assert rc == 0
+
+    cat = Catalog(catalog_dir)
+    patterns = cat.list()
+    assert len(patterns) == 1
+    assert patterns[0].polarity == "problem"
+    # clears the promote floor (freq>=2) but below the default distribution
+    # floor (freq>=3) -> promoted as provisional, not distribution-ready
+    assert patterns[0].status == "provisional"
+    assert patterns[0].distribution_ready is False
+
+    out = capsys.readouterr().out
+    assert "Curate summary" in out
+    # hub offline in tests -> decision queued
+    assert "decisions queued" in out
+
+
+def test_rerun_is_idempotent(tmp_path):
+    cfg_path, store_dir, catalog_dir = _write_config(tmp_path)
+    st = Store(os.path.join(store_dir, "m.db"), os.path.join(store_dir, "blobs"))
+    st.write_digest("claude:a", _digest("claude:a", "claude", "r1", retries=5))
+    st.write_digest("codex:b", _digest("codex:b", "codex", "r2", retries=4))
+    st.close()
+
+    main(["--config", cfg_path, "--auto-approve"])
+    main(["--config", cfg_path, "--auto-approve"])  # second pass: already decided
+    cat = Catalog(catalog_dir)
+    assert len(cat.list()) == 1
+    assert cat.load(cat.list()[0].id).version == "1.0.0"  # no spurious bump
--- a/tests/test_detect_entrypoint.py
+++ b/tests/test_detect_entrypoint.py
@@ -15,6 +15,9 @@ def _digest(uid, flavor, repo, **markers):
        "cost": {"input_tokens": 10, "output_tokens": 1},
        "markers": {"errors": markers.get("errors", 0), "retries": markers.get("retries", 0),
                    "test_runs": 0, "edits": 0, "human_interventions": 0},
+        # fields the quality filter (WP-0005 T01) checks — real coding session
+        "event_count": 40, "first_prompt": "Fix the failing build and retry the suite",
+        "tool_histogram": {"Bash": 20, "Edit": 12, "Read": 8},
    }


--- a/tests/test_detect_quality.py
+++ b/tests/test_detect_quality.py
@@ -0,0 +1,61 @@
+"""Session-quality filter tests (T01)."""
+
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from session_memory.detect.quality import (  # noqa: E402
+    QualityConfig,
+    filter_real,
+    is_real_coding_session,
+    quality_config,
+)
+
+
+def _digest(repo="agentic-resources", events=60, prompt="Implement the curate entrypoint",
+            tools=None):
+    return {
+        "session_uid": "claude:x", "flavor": "claude", "repo": repo,
+        "event_count": events, "first_prompt": prompt,
+        "tool_histogram": tools if tools is not None else {"Bash": 20, "Edit": 15, "Read": 8},
+    }
+
+
+def test_real_session_passes():
+    assert is_real_coding_session(_digest()) is True
+
+
+def test_healthcheck_prompt_dropped():
+    assert is_real_coding_session(_digest(events=3, prompt="Say hello in one word.",
+                                          tools={})) is False
+
+
+def test_interrupted_dropped():
+    assert is_real_coding_session(_digest(events=1, prompt="[Request interrupted by user]",
+                                          tools={})) is False
+
+
+def test_too_short_dropped():
+    assert is_real_coding_session(_digest(events=5)) is False
+
+
+def test_no_repo_dropped():
+    assert is_real_coding_session(_digest(repo=None)) is False
+
+
+def test_no_substantive_tools_dropped():
+    # plenty of events but only plumbing calls -> not real coding
+    assert is_real_coding_session(
+        _digest(tools={"mcp__state-hub__update_task_status": 40})) is False
+
+
+def test_filter_real_keeps_only_real():
+    digs = [_digest(), _digest(events=3, prompt="hello", tools={}), _digest(repo=None)]
+    assert len(filter_real(digs)) == 1
+
+
+def test_quality_config_from_toml():
+    cfg = quality_config({"detect": {"quality": {"min_events": 50}}})
+    assert cfg.min_events == 50
+    assert cfg.min_substantive == 3  # default preserved
--- a/workplans/AGENTIC-WP-0001-statehub-bootstrap.md
+++ b/workplans/AGENTIC-WP-0001-statehub-bootstrap.md
@@ -4,11 +4,11 @@ type: workplan
 title: "Bootstrap State Hub integration"
 domain: helix_forge
 repo: agentic-resources
-status: ready
+status: finished
 owner: codex
 topic_slug: helix-forge
 created: "2026-06-06"
-updated: "2026-06-06"
+updated: "2026-06-07"
 state_hub_workstream_id: "bb9a43a3-a54f-434b-97c2-e1c7142b52f5"
 ---

@@ -20,7 +20,7 @@ Iterating towards optimal agentic performance.

 ```task
 id: AGENTIC-WP-0001-T01
-status: todo
+status: done
 priority: high
 state_hub_task_id: "3ad7b7a9-ee5b-40e7-bd83-d6f0d1db7867"
 ```
@@ -32,7 +32,7 @@ Replace generated placeholders with repo-specific facts where needed.

 ```task
 id: AGENTIC-WP-0001-T02
-status: todo
+status: done
 priority: high
 state_hub_task_id: "db248d57-8e1b-41ef-8386-c04ba490bc6d"
 ```
@@ -45,7 +45,7 @@ changes confidently.

 ```task
 id: AGENTIC-WP-0001-T03
-status: todo
+status: done
 priority: medium
 state_hub_task_id: "9cbb7aa5-dd49-48e2-b5a0-1670584aecf7"
 ```
--- a/workplans/AGENTIC-WP-0004-session-memory-phase2.md
+++ b/workplans/AGENTIC-WP-0004-session-memory-phase2.md
@@ -4,11 +4,11 @@ type: workplan
 title: "Coding Session Memory — Phase 2 (Curate: review workflow + Pattern Catalog)"
 domain: helix_forge
 repo: agentic-resources
-status: ready
+status: finished
 owner: codex
 topic_slug: helix-forge
 created: "2026-06-06"
-updated: "2026-06-06"
+updated: "2026-06-07"
 state_hub_workstream_id: "b3703684-f60e-42f3-b03e-dabe3e8ce3f4"
 ---

@@ -129,7 +129,7 @@ audit trail.

 ```task
 id: AGENTIC-WP-0004-T06
-status: todo
+status: done
 priority: medium
 state_hub_task_id: "95d7747e-8407-41af-9a60-b919a4ee5e06"
 ```
@@ -146,7 +146,7 @@ detect → curate → (Phase 3) distribute flow.

 ```task
 id: AGENTIC-WP-0004-T07
-status: todo
+status: done
 priority: medium
 state_hub_task_id: "20407007-0a8b-4999-a470-fa3c84e17eba"
 ```
@@ -156,10 +156,22 @@ Unit tests for schema/catalog/review/gating on synthetic candidates, plus an
 the live detect output (the Claude+Grok "clean pass" / "abandoned" patterns from
 the WP-0003 verification) into the catalog and confirms a hub decision is logged
 (or queued if the API is down). Confirm catalog round-trips and versioning is
-idempotent on re-run. Refresh design open questions **OQ4/OQ5/OQ6** in
-[DESIGN-session-memory.md](../docs/DESIGN-session-memory.md). After workplan file
-updates, notify the custodian operator to run from `~/state-hub`:
+idempotent on re-run. Refresh design open questions **OQ4/OQ5/OQ6** (PRD §12).
+After workplan file updates, notify the custodian operator to run from
+`~/state-hub`:

 ```bash
 make fix-consistency REPO=agentic-resources
 ```
+
+**Verification results (2026-06-07):** full suite 72/72 green (26 new curate
+tests across schema/catalog/review/gating/decisions/entrypoint). Live pipeline
+over real local sessions: fresh ingest 94→93 → 72 digests; detect surfaced 3
+candidates, **2 cross-flavor** (Claude+Grok). `curate --auto-approve` promoted
+all 3 into the files-first catalog — `sp-success-clean_pass-outcome` and
+`sp-problem-abandoned-outcome` (both cross-flavor, `approved`/`distribution_ready`)
+plus `sp-problem-budget_overrun-tokens` (Claude-only). 3 hub decisions queued
+(API offline). Re-run was fully idempotent (3 skipped, 0 catalog writes, no
+version bump). PRD §12 OQ4/OQ5/OQ6 resolved. The 3 catalog artifacts are
+committed as the source of truth; operator runs `make fix-consistency` to index
+them in the hub.
--- a/workplans/AGENTIC-WP-0005-detect-hardening.md
+++ b/workplans/AGENTIC-WP-0005-detect-hardening.md
@@ -0,0 +1,88 @@
+---
+id: AGENTIC-WP-0005
+type: workplan
+title: "Coding Session Memory — Detect Hardening (quality filter + infra signals)"
+domain: helix_forge
+repo: agentic-resources
+status: ready
+owner: codex
+topic_slug: helix-forge
+created: "2026-06-07"
+updated: "2026-06-07"
+state_hub_workstream_id: "d8b7b8d1-1d85-4d2a-8ccd-7b0366a9442d"
+---
+
+# Coding Session Memory — Detect Hardening
+
+A focused hardening pass (call it Phase 1.5) so the Detect output is trustworthy
+enough to drive an **infrastructure assessment**. Triggered by ad-hoc analysis of
+the live store after Phase 2:
+
+- Of **72 captured sessions, only 31 are real coding sessions**; the rest are
+  health-checks / smoke-tests / interrupted runs (mostly `llm-connect` *"Say hello
+  in one word"*). The `abandoned` outcome heuristic mislabels these, and Phase 2
+  cataloged a **false-positive** "cross-flavor abandoned" pattern as
+  `approved`/`distribution_ready`.
+- All 31 real sessions read as `success`, so the current signal set
+  (outcome + markers + cost) surfaces almost no genuine friction.
+- The already-captured `tool_histogram` tells the real story: **~17% of tool
+  activity in real sessions is State Hub MCP + task plumbing + `ToolSearch`
+  schema-loading**, concentrated to 40–70% in some sessions — but `signals.py`
+  never looks at it.
+
+No new capture is needed — this is analysis the data already supports.
+
+## Session-Quality Filter
+
+```task
+id: AGENTIC-WP-0005-T01
+status: done
+priority: high
+state_hub_task_id: "9f8b4304-0a37-4f66-ad34-d93e12fba0d8"
+```
+
+Add `detect/quality.py` with `is_real_coding_session(digest)` that filters out
+health-checks, smoke-tests, interrupted, and trivially-short sessions (event-count
+floor, repo present, substantive edit/tool activity, not a single hello/interrupt
+prompt). Wire it into the detect pipeline so signals/clusters only form over real
+sessions — fixing the `abandoned` false-positive. Knobs under `[detect]` in
+`config.toml`. Unit-tested on synthetic trivial-vs-real digests.
+
+## Infra-Overhead + Thrash Signals
+
+```task
+id: AGENTIC-WP-0005-T02
+status: todo
+priority: high
+state_hub_task_id: "10d57b05-a731-4ece-bf45-f6a98ac77555"
+```
+
+Add `tool_histogram`-based extractors to `detect/signals.py`: a shared tool-bucket
+helper (`shell` / `edit` / `read` / `statehub_mcp` / `task_mgmt` / `schema_load` /
+`other`); `sig_infra_overhead` (PROBLEM when the statehub+task+schema share of tool
+calls exceeds a threshold; magnitude = share; locus `infra_overhead`);
+`sig_schema_thrash` (`ToolSearch` count over threshold; locus `schema_load`);
+`sig_tool_thrash` (extreme single-tool repetition). Pure functions over digests;
+thresholds configurable. Unit-tested.
+
+## Re-run Live, Purge False Positives, Ranked Friction Report
+
+```task
+id: AGENTIC-WP-0005-T03
+status: todo
+priority: high
+state_hub_task_id: "8b9d029a-60d0-4caf-af62-4fcc9c9a645c"
+```
+
+Re-run `ingest → detect` over the real local sessions with the filter + new
+signals. Purge the false-positive catalog entries seeded in Phase 2 (the
+health-check `abandoned` pattern) and re-curate so the catalog reflects real
+friction. Produce a ranked **friction assessment** (`docs/ASSESSMENT-infra-friction.md`)
+of the major infrastructure problems — quantified per repo/flavor, infra-overhead
+share, schema-thrash — with recommendations (incl. the State Hub / MCP skill
+hypothesis). After workplan file updates, notify the operator to run from
+`~/state-hub`:
+
+```bash
+make fix-consistency REPO=agentic-resources
+```
Author	SHA1	Message	Date
tegwick	70433cda61	session-memory: session-quality filter (WP-0005 T01) detect/quality.py: is_real_coding_session drops health-checks / smoke-tests / interrupted / trivially-short sessions (event floor, repo present, substantive tool activity, non-trivial prompt). Wired into run_detect so signals only form over real sessions — fixes the abandoned false-positive. [detect.quality] knobs; existing detect/curate fixtures made realistic. 8 new tests; suite 80/80. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-07 11:07:22 +02:00
tegwick	56b2f576de	AGENTIC-WP-0001: complete T02 + close bootstrap workplan T02 was the one genuinely-incomplete bootstrap task: AGENTS.md had no dev-workflow section. Added one documenting the pure-stdlib Python 3.11+ toolchain, pytest, and the session_memory ingest/detect/curate entrypoints so future sessions can verify changes. T01 (integration files) and T03 (first real workplan) were already satisfied; reconciled stale ready/todo bookkeeping to finished/done. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-07 10:15:23 +02:00
tegwick	d06791f070	session-memory Phase 2: verify + catalog artifacts (T07) End-to-end verification over real local sessions: ingest 94->93 -> 72 digests; detect 3 candidates (2 cross-flavor); curate --auto-approve cataloged 3 SolutionPatterns (2 cross-flavor approved/distribution_ready, 1 Claude-only), re-run fully idempotent, 3 hub decisions queued (API offline). Commits the 3 catalog artifacts as the source of truth. PRD §12 OQ4/OQ5/OQ6 marked resolved; README + design refreshed. Workplan finished; suite 72/72. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-07 10:08:52 +02:00
tegwick	519e76442a	session-memory Phase 2: curate entrypoint + README (T06) python -m session_memory.curate: refreshes detect candidates, then drives them through review interactively or with --auto-approve (batch, gate-driven) / --json. Emits a catalog diff summary; queues hub decisions when offline. [curate] config gains decision_queue + workstream id. README documents the detect -> curate -> distribute flow and the gate knobs. 2 new tests; suite 72/72. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-07 10:00:56 +02:00