@bongos/core 1.19.549 → 1.19.551

package/.bongos-core.json

@@ -2,22 +2,22 @@
   "artifact": "bongos-core",
   "manifest_schema": 1,
   "generator": "scripts/gds/package-core.js",
-  "core_version": "1.19.549",
-  "core_contract": "1.19.549",
-  "source_commit": "a554c51b4ea9134b53e0f99617a051c65cf1fb34",
+  "core_version": "1.19.551",
+  "core_contract": "1.19.551",
+  "source_commit": "de0b6c99753fd502f2356bcfb6bf338227445cc7",
   "source_ref": "HEAD",
-  "built_at": "2026-09-04T18:37:41.417Z",
+  "built_at": "2026-09-04T20:39:18.276Z",
   "redaction": {
     "model": "docs-redacted+functional-verbatim",
-    "docs_redacted": 441,
+    "docs_redacted": 442,
     "agent_docs_stubbed": 24,
-    "functional_verbatim": 2013,
+    "functional_verbatim": 2018,
     "rules": 3,
     "gate_literals": 3,
     "gate": "passed"
   },
-  "file_count": 2478,
-  "tree_sha256": "9cee93593f3443ba78ae1c00bb5cd092008ed33ee8c024f4e61a7d17d67bffb5",
+  "file_count": 2484,
+  "tree_sha256": "38fea0008aa898d10d6e7388c30a7d422be81dcde8d201172848e551d72d361b",
   "files": [
     {
       "path": ".claude/skills/blocker-review/SKILL.md",
@@ -29,6 +29,11 @@
       "mode": "0000644",
       "sha256": "90de8b32fb4dac69d6e5c781d78489c36c40f00c4b07f062c0650cff08052203"
     },
+    {
+      "path": ".claude/skills/bug-triage/SKILL.md",
+      "mode": "0000644",
+      "sha256": "85481aea23304149520f1adcb7e2e8ab05b655c269dd2e38a6c04922013e9202"
+    },
     {
       "path": ".claude/skills/builder-backup/SKILL.md",
       "mode": "0000644",
@@ -177,7 +182,7 @@
     {
       "path": ".claude/skills/idea-triage/SKILL.md",
       "mode": "0000644",
-      "sha256": "b64fe42bb2a542304802b48bea348ae817608ac47f026d98c4de9bea5aa90174"
+      "sha256": "2519755269ff2d55e3c060569f7d16c440428ba005d206a3bf12fe2275830285"
     },
     {
       "path": ".claude/skills/ideate/SKILL.md",
@@ -367,12 +372,12 @@
     {
       "path": "clients/bongos-client/README.md",
       "mode": "0000644",
-      "sha256": "b1abf8c13f89ba8fe2ebca540d1a8b22fac0ad293595e51f464e189dfaf1a1ad"
+      "sha256": "262491824b5d7b6005c69d8190d50f450320f02dc1333dc7ae029616aff9f018"
     },
     {
       "path": "clients/bongos-client/bongos-client.global.js",
       "mode": "0000644",
-      "sha256": "1178634ee16c7b1a83defe0c6ece5754a92901370ce18f114a84d3922f7065ca"
+      "sha256": "fd5d716ec3ba7fbdb9fdaee076ef89048424248139ef7d4ea51c3983866dea90"
     },
     {
       "path": "clients/bongos-client/examples/hello-world.mjs",
@@ -382,17 +387,17 @@
     {
       "path": "clients/bongos-client/index.cjs",
       "mode": "0000644",
-      "sha256": "be9e267735c41c695157c01fb82efb9219c2f6769febfa46495395d6dcc887ec"
+      "sha256": "17753bb744e592f7d7871d33399536d9938b1b7d6c6249959bcbb4355688b557"
     },
     {
       "path": "clients/bongos-client/index.d.ts",
       "mode": "0000644",
-      "sha256": "73e1e22f18262eb0cb73edfe58aef773f802bae33b68b9ca2e4430403d03d7c9"
+      "sha256": "1b28c414b039d21ab8105d1687cf6d75cde078238ac84b920885772cd5e9bdbf"
     },
     {
       "path": "clients/bongos-client/index.mjs",
       "mode": "0000644",
-      "sha256": "f9c26f336f8b7808ed1c07e844191abd0de7bc39657dda5d0363f5d472b6c2c4"
+      "sha256": "779dc3d8970036513a0f8da8e4846bd6c0d2d4714de38390b14ede249e75a3f6"
     },
     {
       "path": "clients/bongos-client/package.json",
@@ -1767,12 +1772,12 @@
     {
       "path": "docs/api-reference.md",
       "mode": "0000644",
-      "sha256": "aef38c3ecd8942f0a38c56cbb390fd3590e785634d24e95660900a5140636b5f"
+      "sha256": "5ec1ca7cb551994a31d644eeebd77795955bac39e6389feaf6e6a5ec0217c911"
     },
     {
       "path": "docs/api/openapi.json",
       "mode": "0000644",
-      "sha256": "923202b89890cc048769c8b05087181fa06ff126aa80aeaa4d9f669f4d96c27e"
+      "sha256": "40d863b7b46cc3691298f9e48c254cda3b7f4ab23bb008712a55e75764a69dd1"
     },
     {
       "path": "docs/architecture.md",
@@ -2637,7 +2642,7 @@
     {
       "path": "docs/file-map.md",
       "mode": "0000644",
-      "sha256": "6085fdda80f0c961067a044b890358fa148415638854029e63f764e262594aeb"
+      "sha256": "e7c193b1cca2fc3f59742e95f921fc65096a32fbbcfba16e4a3950448293a7ba"
     },
     {
       "path": "docs/handoff-template.md",
@@ -2647,7 +2652,7 @@
     {
       "path": "docs/module-api-changelog.md",
       "mode": "0000644",
-      "sha256": "c4f07821d5cae7bacdc51e2d7a0f0c5672beb7ba623069c9b4f88e565d8f77b6"
+      "sha256": "1389ff20919a19aef2fbff28102f596f0ca168c34bc4c4114cf557bbf52627a5"
     },
     {
       "path": "docs/modules-contract.md",
@@ -3644,6 +3649,11 @@
       "mode": "0000644",
       "sha256": "8ae32be783a4ca41637f56626bd83375aa4ae0755e701a0c0170eada8d8303b8"
     },
+    {
+      "path": "migrations/core_232_task_merge_provenance.sql",
+      "mode": "0000644",
+      "sha256": "5ed2ff8117e6650de621a1576062b27f2d6e84faf8860f9cb2aed45323a85aa6"
+    },
     {
       "path": "modules/agents/lib/validate.js",
       "mode": "0000644",
@@ -4042,7 +4052,7 @@
     {
       "path": "modules/dev-box/app/src/vendor/bongos-client.cjs",
       "mode": "0000644",
-      "sha256": "be9e267735c41c695157c01fb82efb9219c2f6769febfa46495395d6dcc887ec"
+      "sha256": "17753bb744e592f7d7871d33399536d9938b1b7d6c6249959bcbb4355688b557"
     },
     {
       "path": "modules/dev-box/app/test/unit.mjs",
@@ -4282,7 +4292,7 @@
     {
       "path": "modules/government/catalog.js",
       "mode": "0000644",
-      "sha256": "913b7ab33c6785903376b0bf7804d7d8d2db620112b983ae4c0368f33a6b8c0f"
+      "sha256": "57b29298fa5707d3a1d59e78612d2b039b1559d26b8a9116812d5c76e9c3af87"
     },
     {
       "path": "modules/government/config.js",
@@ -4384,6 +4394,11 @@
       "mode": "0000644",
       "sha256": "2b729acbc94c933ff2558d59e3cb811e7725450f0d646f9d9bc89209c1340d0f"
     },
+    {
+      "path": "modules/government/migrations/government_011_task_merge_metic.sql",
+      "mode": "0000644",
+      "sha256": "ccd2cd14103bae00957227d037d41f76adaaddc639bd09808f244b0bc514e702"
+    },
     {
       "path": "modules/government/module.json",
       "mode": "0000644",
@@ -5437,7 +5452,7 @@
     {
       "path": "modules/lifecycle/db-ship.js",
       "mode": "0000644",
-      "sha256": "d300392b7bf802d00ae79c55ad3b0d9413dc9bb56e7e1105f49656b39abb874a"
+      "sha256": "e48d877a5951f02153c7384217421b2b04a3a1fed215b61ac41fc5b0b8355ded"
     },
     {
       "path": "modules/lifecycle/db-tasks.js",
@@ -5452,7 +5467,7 @@
     {
       "path": "modules/lifecycle/db.js",
       "mode": "0000644",
-      "sha256": "09d40dd2eaae47c4620993ad3db6a6db2d880cc499724d31ceff29604960fd64"
+      "sha256": "df5de732d9484e1bcf801c657c78b618d1a0353a912cdb1ee4fdd2c9f121749c"
     },
     {
       "path": "modules/lifecycle/dead-deps.js",
@@ -5682,7 +5697,7 @@
     {
       "path": "modules/lifecycle/routes/tasks.js",
       "mode": "0000644",
-      "sha256": "0ba020342e32eb552532a3a34a57b19aa4bbe87f0cb89ace0c45f4c301d7a015"
+      "sha256": "366cb9c60952539ab8a946f8cf757754163b6cd6764a6430777b4529c86c6a75"
     },
     {
       "path": "modules/lifecycle/routes/versions.js",
@@ -7452,12 +7467,12 @@
     {
       "path": "package-lock.json",
       "mode": "0000644",
-      "sha256": "8a3f8bc746c2704c4859a36a3f365d6c24305019ba6270aa2c25abbd6bba1bef"
+      "sha256": "c71d59c13b87d9ffbd6c792eb56b26eb434704e855aed3afdfc58914dba02e29"
     },
     {
       "path": "package.json",
       "mode": "0000644",
-      "sha256": "2a7d837ba22c70e21b083f3189108a90a974312b07bccf634bca4175e7213697"
+      "sha256": "6a2c55e98a36d257177254a100569a14bec9b3720931caf9ef265971b81ba319"
     },
     {
       "path": "public-docs/index.html",
@@ -7644,6 +7659,11 @@
       "mode": "0000644",
       "sha256": "afcebfb8e8e8f7041ee9cbfce1e9b8a46a299feb5e6e89ce4e00273135b268db"
     },
+    {
+      "path": "scripts/gds/bug-triage.js",
+      "mode": "0000644",
+      "sha256": "70b8f34cd401395e828bcd7d2b1c3bf508d2f1e0918f998df034efb15b6783b8"
+    },
     {
       "path": "scripts/gds/build-bongos.js",
       "mode": "0000644",
@@ -8702,7 +8722,7 @@
     {
       "path": "scripts/gds/triage.js",
       "mode": "0000644",
-      "sha256": "831de1914642a7dcf7c591c9cc69b37c917122e04428cb1e28fc00b07170b603"
+      "sha256": "dc494030979bf60de1d1423c68a07e36667e816d7e2a1652960cfbd17be4bfcc"
     },
     {
       "path": "scripts/gds/turn-end-sound.js",
@@ -9147,7 +9167,7 @@
     {
       "path": "src/module-api.js",
       "mode": "0000644",
-      "sha256": "9808388b3f47f916911b59d0434a3e67f466869be9a3c7fa9adde9fb1274770c"
+      "sha256": "95e7b476152291d84e955bfd585e62c48c48d85dcab0ca8d73c4519af2997e8b"
     },
     {
       "path": "src/module-loader/catalog.js",
@@ -9197,7 +9217,7 @@
     {
       "path": "tests/abandon_task.mjs",
       "mode": "0000644",
-      "sha256": "031b75aeb83322055c4d69ab5481b220a7a41ccbe950129b84f8bc80d7c75a40"
+      "sha256": "ccfb5d3e85fd6bd34acc78d510edf20d037d344eca58706f3e3130fa6dce611c"
     },
     {
       "path": "tests/access_requests_invite.mjs",
@@ -9599,6 +9619,11 @@
       "mode": "0000644",
       "sha256": "b9303554061efaea510ef42d774ab9017ac7258cc0bc08f684abca7408b3fa86"
     },
+    {
+      "path": "tests/bug_triage.mjs",
+      "mode": "0000644",
+      "sha256": "2bc6312ff72fffc04478d40adc60e57e031b52ed208adf1d88f78986e536623f"
+    },
     {
       "path": "tests/build_info.mjs",
       "mode": "0000644",
@@ -12214,6 +12239,11 @@
       "mode": "0000644",
       "sha256": "bc89502f571659477178d68047098cf6098030491ac5c5a82b46802474a48387"
     },
+    {
+      "path": "tests/task_merge.mjs",
+      "mode": "0000644",
+      "sha256": "2a16f120729c641de65cccf79f1583e2d2e68b72bf10277310991c1787dde2bf"
+    },
     {
       "path": "tests/task_ownership_gate.mjs",
       "mode": "0000644",
@@ -12302,7 +12332,7 @@
     {
       "path": "tests/triage_development.mjs",
       "mode": "0000644",
-      "sha256": "74798ce81280d4fcfa5844b575c3c64db582308f7a0726fc420317fe83eadd96"
+      "sha256": "56ff1fab014462291349ac200aab9d12e11a73c028aebeb564866b7a1b309fbd"
     },
     {
       "path": "tests/ui_design_kit.mjs",

package/.claude/skills/bug-triage/SKILL.md

@@ -0,0 +1,191 @@
+---
+name: bug-triage
+description: >-
+  Daily walk through open kind=bug tasks, which land with no triage gate and
+  accumulate near-duplicates. Groups likely dups for one batch verdict, then
+  walks what's left: merge into a canonical task or won't-fix. Mirrors
+  /idea-triage. Triggers on "/bug-triage", "triage bugs", "walk the bug
+  queue", "dedupe bugs". Metic+ rank only (not Xenos/Thetes).
+---
+
+You are running the bug-triage session for Example. Open `kind=bug` tasks get
+walked and verdicted, so near-duplicate reports don't sit forever as separate
+claimable rows.
+
+## Why this exists (task 1001478, idea 244)
+
+Since task 1003071, a `#bugs` report routes straight to a `kind=bug` task at
+filing time — no inbox stop, no gate. That is correct for getting a report
+into the queue fast, but it means nothing ever folds two reports of the same
+underlying bug together the way `/idea-triage` folds duplicate ideas. This
+skill is that missing step, built on a **task-merge primitive** (`tasks`
+gained `merged_into_task_id` + `triage_note` in `core_232`, mirroring
+`idea_inbox`'s `merged_into_idea_id` from `core_195`) that did not exist
+before this task.
+
+**A merge is `abandoned` + a target, not a new status.** `tasks.status` has no
+`'merged'` value — adding one would touch every place that already enumerates
+task statuses (the claimable feed, dependency satisfaction, the diagrams
+state machine) to answer a question `abandoned` already answers: this task is
+retired, and by owner ruling its dependents are stranded, not waiting (see
+`modules/lifecycle/CLAUDE.md`). `merged_into_task_id` carries the one thing
+bare abandon can't: which task absorbed it. Read `modules/lifecycle/db-ship.js`
+`mergeTask`'s header comment for the full guard list before touching this
+surface.
+
+## Rank gate — Metic+ only
+
+**Gated to Metic+ (Metic, Archon); not Xenos or Thetes.** Merging or
+won't-fixing a bug task reshapes the team's backlog the same way idea triage
+does — `task.merge` and `task.abandon` are both floor-`metic` permissions
+(`modules/government/catalog.js`). A sub-Metic can still **file** a bug (via
+Discord `#bugs`, or `POST /tasks` if `task.create` allows it); they can't
+verdict the queue.
+
+**Enforce the gate at the very top of the session, before reading anything else:**
+
+1. Call `GET /api/gds/me` and read `builder.rank`. Compare case-insensitively.
+2. **If `rank` lowercases to `xenos` or `thetes`** → stop immediately. Tell the
+   builder: *"Bug triage is a Metic+ operation — merging or dismissing a bug
+   report reshapes the team's backlog. File new bugs via Discord `#bugs`;
+   ask an Archon to promote you if you need to triage them."* Do not list
+   open bugs, do not propose verdicts.
+3. **If `rank` lowercases to `metic` or `archon`** → proceed below.
+4. **If `rank` is absent** (pre-rank deployment) → treat the session as
+   available, the same pre-rank passthrough `/idea-triage` uses.
+
+> Belt-and-braces: the server already rejects `POST /tasks/:id/merge` and
+> `POST /tasks/:id/abandon` with `403 permission_forbidden` for a sub-Metic
+> caller — this check just stops one early instead of after a wall of 403s.
+
+## What this skill does
+
+1. Lists open `kind=bug` tasks via `GET /api/gds/tasks?kind=bug&limit=1000`,
+   filtered to the still-open statuses (`backlog`, `ready`, `active`,
+   `blocked`, `completed`, `confirmed` — never `shipped` or `abandoned`,
+   the latter of which already includes merged-away duplicates).
+2. **Batches near-duplicate bug tasks** into one verdict per cluster
+   (`node scripts/gds/bug-triage.js`'s reuse of `triage.js`'s `clusterRecords`
+   — the same union-find/Jaccard core `/idea-triage`'s `clusterIdeas` uses,
+   generalized in task 1001478 so the algorithm is written once). Partitioned
+   by `goal_id`, not `kind` — every row here is already `kind='bug'`, so a
+   kind gate would be a no-op; two bug reports in unrelated goals describing
+   similar-sounding symptoms are not the same triage decision.
+3. For each remaining bug, prompts the user with: **merge** (this is the same
+   report as task N) or **won't-fix** (not a bug / already fixed / can't
+   reproduce — closes the report itself, folds into nothing).
+4. Applies each verdict via `POST /api/gds/tasks/:id/merge` or
+   `POST /api/gds/tasks/:id/abandon`.
+
+**No "promote" step, unlike `/idea-triage`.** A bug report is already a task —
+there is no shape question (goal vs criterion vs task) to work out, because
+`task-classifier.js`'s auto-routing already decided `kind=bug` at capture
+time. The only two questions triage adds are "is this the same as an existing
+report" and "should this report be closed at all".
+
+## How to use
+
+**Step 1 — fetch open bug tasks.**
+
+```
+node scripts/gds/api.js GET "/api/gds/tasks?kind=bug&limit=1000"
+```
+
+Filter client-side to the open statuses above. If nothing remains, print "bug
+queue empty — nothing to triage today" and stop.
+
+For a human at a terminal, `node scripts/gds/bug-triage.js` runs the whole
+walk interactively instead — batching, per-task prompts, and the summary all
+in one command, same shape as `node scripts/gds/triage.js`.
+
+**Step 1.5 — batch near-duplicate bug reports BEFORE walking individually.**
+
+Run `clusterRecords` (`scripts/gds/triage.js`, pure, no I/O) with
+`textOf: t => t.title + ' ' + t.description`, `kindOf: t => t.goal_id`,
+`timeOf: t => new Date(t.created_at)`. For each cluster it returns, offer the
+user ONE decision covering the whole group:
+
+> Cluster of 3 likely duplicates: #501 "Login button does nothing on
+> Safari", #517 "Safari: login unresponsive", #530 "Login broken in Safari".
+> **Keep #501 and merge the rest, dismiss all as won't-fix, walk them
+> individually, or skip?**
+
+- **Keep #N** → merge every other member into #N (`POST /tasks/:id/merge
+  {"into_task_id": N}`), leaving #N itself open — it still gets the normal
+  Step 2 walk below, since "which report is the real one" and "should it stay
+  open" are different decisions.
+- **Won't-fix all** → every member abandoned with one shared reason.
+- **Individually** → ignore the grouping for this cluster; each member
+  returns to the normal walk.
+- **Skip** → leave the whole cluster open; it clusters again next triage.
+
+`node scripts/gds/bug-triage.js` already runs this pass automatically before
+its interactive walk — a human running the deterministic walker gets it for
+free with no flag.
+
+**Step 2 — walk the list, task by task.**
+
+Present each remaining bug with:
+- id, status, title
+- description (full)
+- goal_id, created_at
+
+Then ask the user (concise prompt — the owner may not be a software engineer,
+CLAUDE.md §2's portable default; frame in product/outcome terms, not
+implementation):
+
+> Bug #N: "{title}". {one-sentence summary of the report}. **Is this a
+> duplicate of an existing bug, should it be closed as won't-fix, or should
+> it stay open?**
+
+**Step 3 — apply the verdict.**
+
+**Merge** → ask which existing task N is the canonical one, then fold this
+report into it:
+
+```
+node scripts/gds/api.js POST /api/gds/tasks/<id>/merge --body '{"into_task_id":<N>,"reason":"same report as #N"}'
+```
+
+The server refuses a self-merge (`merge_self`), a shipped source
+(`cannot_merge_shipped`), a task with an active claim
+(`task_has_active_claim`), a target that doesn't exist
+(`merge_target_not_found`), and a target that is itself merged elsewhere
+(`merge_target_merged` — it names where the chain ends; retarget there
+instead of re-merging).
+
+**Won't-fix** → dismiss with a reason (not a bug / already fixed / cannot
+reproduce):
+
+```
+node scripts/gds/api.js POST /api/gds/tasks/<id>/abandon --body '{"reason":"cannot reproduce on current build"}'
+```
+
+**Keep open** is implicit — leave the task un-patched by not calling either
+route. It will surface again next triage.
+
+**Step 4 — surface a summary.**
+
+Report to the user:
+- Total walked
+- Merged (count + which task each folded into)
+- Won't-fixed (count + ids)
+- Skipped / left open (count — these surface again next session)
+- Remaining open bug count
+
+## Why we run this
+
+The same reason `/idea-triage` runs daily: a queue nobody walks quietly fills
+with near-duplicates that each cost a separate builder's attention to
+discover independently. A fast pass with mostly "skip" verdicts beats
+skipping the session entirely — the cadence is the discipline.
+
+## Tone for the user
+
+The owner is the prompter (CLAUDE.md §2's portable default). Frame merge/
+won't-fix decisions in terms of what's actually broken for the end user, not
+implementation: "these
+are the same login failure reported twice" not "these share 71% word
+overlap." Decide the mechanical merge target yourself; ask the human only
+when two reports are genuinely ambiguous (different symptoms that might
+share one root cause, say) rather than obviously the same.

package/.claude/skills/idea-triage/SKILL.md

@@ -53,9 +53,10 @@ capture-time suggester (R08/R09/R11 surfaces) is where the fix belongs.
 ## What this skill does
 
 1. Lists open ideas via `GET /api/gds/inbox`.
-2. **Develops the open ideas in parallel** — spawns one subagent per idea (the Agent tool) that fleshes the one-liner out into a structured *open-questions / challenges / risks* block, so the Archon verdicts on a developed idea instead of a bare title (task 930).
-3. For each, prompts the user with: **promote** (create a task and link it), **discard** (closed, no action), or **merge** (this duplicates idea N, mark merged into N).
-4. Updates each idea via `PATCH /api/gds/inbox/:id` with the verdict.
+2. **Batches near-duplicate ideas** into one verdict per cluster (`node scripts/gds/triage.js`'s `clusterIdeas`), so a backlog of similar asks doesn't cost one decision each (task 1001465).
+3. **Develops what's left with a LEAN inline pass by default** — reasoning from context already in the session, no subagent spend — and escalates to a FEW targeted subagents only where repo grounding would actually flip the verdict (task 1001465; the old unconditional 12-subagent fan-out is now opt-in, task 930's original mechanism).
+4. For each, prompts the user with: **promote** (create a task and link it), **discard** (closed, no action), or **merge** (this duplicates idea N, mark merged into N).
+5. Updates each idea via `PATCH /api/gds/inbox/:id` with the verdict.
 
 ## How to use
 
@@ -91,33 +92,46 @@ this order and prints the grade on each card, and `render-ideas.js` renders the
 same two sections — so the interactive walk, the generated file and this skill all
 show the same idea the same way.
 
-**Step 1.5 — develop the open ideas in parallel (task 930).**
+**Step 1.4 — a cheap staleness pass, before anything else costs a token.** Quick-skim each open idea's title/body against the current codebase/CLAUDE.md to catch ones already implemented, already superseded, or otherwise obviously stale. A recent triage found 47 of 170 open ideas already stale — filtering those first means neither the batching pass nor the development pass below is spent on an idea that just needs a discard.
 
-**Before spawning any subagents, run a cheap staleness pass over the list first** — a quick skim of each idea's title/body against the current codebase/CLAUDE.md to catch ideas that are already implemented, already superseded, or otherwise obviously stale. A recent triage found 47 of 170 open ideas already stale; filtering those out first means the expensive multi-agent development pass below is only spent on ideas that still need it.
+**Step 1.5a — batch near-duplicate ideas BEFORE developing anything (task 1001465).**
 
-Then, before walking the list, **flesh out each remaining open idea with a subagent** so the verdict lands on a developed idea, not a one-liner. Spawn the subagents **in parallel** (multiple `Agent` tool calls in a single message) — they run while you prepare the walk, so there's no serial wait.
+Run `clusterIdeas` (`scripts/gds/triage.js` — pure, no I/O): it groups open quick/spark ideas that share a `kind` and a high title/body word overlap. For each cluster it returns, offer the user ONE decision covering the whole group instead of walking each member — this is the actual fix for "triage costs a ton of credits" (idea 377): a backlog of near-identical asks used to cost one full development + verdict PER idea, and now costs one:
 
-- **Model:** `opus` for every development subagent. This is research / design analysis against the live repo, which per CLAUDE.md §2's routing table is Opus work — never inherit silently, never downgrade to Haiku.
-- **Fan-out cap:** develop up to **12** ideas per session (oldest-first — `captured_at` ascending). If there are more than 12 open, note "developed the 12 oldest; the rest are walked undeveloped" so nothing is silently dropped.
-- **One idea per subagent.** Give each subagent the idea's `id`, `title`, `body_md`, `kind`, and `captured_by`, and this instruction:
+> Cluster of 3 likely duplicates: #101 "Add dark mode toggle", #145 "Dark mode support", #203 "Support a dark theme". **Keep #101 and merge the rest, discard all, walk them individually, or skip?**
 
-  > Flesh out this Example idea for an Archon triage verdict. Read the relevant code/docs to ground yourself (start from CLAUDE.md and the surfaces the idea touches). Return ONLY a JSON object with these keys: `summary` (one sentence, what integrating it actually means), `integration_surfaces` (array of files/areas it would touch), `open_questions` (array — the decisions a human must make before this can be a task), `challenges` (array — what's hard or unclear), `risks` (array — what could break or regress), `rough_effort` (one of `S`/`M`/`L`), `recommendation` (one of `promote`, `discard`, `merge:#<id>`, `needs-info`). Keep each array item to one line. Do not invent surfaces you didn't verify exist.
+- **Keep #N** → merge every other member into #N (`merged_into_idea_id`), leaving #N itself open — it still gets the normal Step 2 walk below, since "which of these duplicates is real" and "should the real one become a task" are different decisions.
+- **Discard all** → every member closed, one shared `triage_note`.
+- **Individually** → ignore the grouping for this cluster; each member returns to the normal walk.
+- **Skip** → leave the whole cluster open; it clusters again next triage.
 
-- **Collect the results into a development cache** keyed by idea id and write it to a temp file so the deterministic walker can also surface it:
+`node scripts/gds/triage.js` already runs this pass automatically before its interactive walk — if a human is running the deterministic walker instead of you, they get it for free with no flag.
+
+**Step 1.5b — develop what's LEFT with a lean inline pass by default (task 1001465).**
+
+**The subagent fan-out below is now OPT-IN, not the default.** The prior design ran a 12-Opus-subagent fan-out unconditionally on every session — the main cost driver idea 377 was filed about. Default instead to a **lean inline pass**: for each remaining open idea (after batching), reason about it yourself using context you already have — CLAUDE.md, `/recall`, a quick grep if genuinely needed — and write a short inline note (one or two lines: what it touches, the one open question that matters) directly in the session. No `Agent` tool call, no JSON cache file, no per-idea cost.
+
+**Escalate to a targeted subagent only when the inline pass can't tell whether repo grounding would flip the verdict** — a technical/GDS idea where "is this already implemented?", "does this surface still exist?", or "would this conflict with X" needs an actual code read to answer, not a guess. Cap this at a **FEW** ideas per session (3, not 12) — if more than 3 are genuinely ambiguous, develop the 3 oldest and note the rest are walked on the inline pass alone.
+
+- **Model:** `opus` for a targeted subagent — this is still research/design analysis against the live repo (CLAUDE.md §2's routing table), never downgrade to Haiku just because there are fewer of them now.
+- **One idea per subagent**, same instruction as before: give it the idea's `id`, `title`, `body_md`, `kind`, `captured_by`, and ask for the same JSON shape (`summary`, `integration_surfaces`, `open_questions`, `challenges`, `risks`, `rough_effort`, `recommendation`) — only the FAN-OUT WIDTH changed, not the shape a developed idea returns.
+- **Collect results into the same development cache**, written to `${TMPDIR:-/tmp}/otb-idea-development.json` (Windows: `%TEMP%\otb-idea-development.json`):
 
   ```json
   { "generated_at": "<ISO>", "ideas": { "228": { "summary": "...", "integration_surfaces": ["..."], "open_questions": ["..."], "challenges": ["..."], "risks": ["..."], "rough_effort": "M", "recommendation": "needs-info" } } }
   ```
 
-  Write it to `${TMPDIR:-/tmp}/otb-idea-development.json` (Windows: `%TEMP%\otb-idea-development.json`). A subagent that fails or returns unparseable JSON is simply omitted from the cache — that idea is walked undeveloped, not blocked.
+  A subagent that fails or returns unparseable JSON is simply omitted — that idea is walked on the inline pass alone, not blocked.
+
+**The full 12-idea fan-out is still available, explicitly, when the queue genuinely needs it** — a big backlog day, or the user asks for "the deep pass" / "develop all of these with subagents". When asked for it: spawn one Opus subagent per open idea (not just the ambiguous few), up to the same 12-idea cap (oldest-first), same instruction and cache shape as above. This is the OLD default, now a deliberate choice rather than something that runs unasked every time.
 
-If a human will run the interactive walker instead of the Claude-driven walk below, hand it the cache:
+If a human will run the interactive walker instead of the Claude-driven walk below, hand it the cache (works the same whether it came from the lean pass's few subagents or the full deep pass):
 
 ```
 node scripts/gds/triage.js --development "${TMPDIR:-/tmp}/otb-idea-development.json"
 ```
 
-`triage.js` prints each idea's developed block above its verdict prompt. It never *generates* the analysis (that's this skill's subagent fan-out) — it only displays what the cache holds, and runs exactly as before when no cache is present.
+`triage.js` prints each idea's developed block above its verdict prompt, and runs its own batch-cluster pass (Step 1.5a) automatically first. It never *generates* the analysis (that's this skill's job); it only displays what the cache holds, and runs exactly as before when no cache is present.
 
 > **Not in this cut (follow-up leg).** The next step is to let the Archon push the developed `open_questions` *back to the submitter* — a round-trip in the Discord `#ideas` thread / DM before promotion — so the submitter says how they'd want it integrated. That needs the Discord bot interaction + idea↔thread linking (idea `#216` / [ADR 0033](../../../docs/adr/<redacted>.md)) and is tracked separately. For now the open questions are answered by the Archon in-session.
 

package/clients/bongos-client/README.md

@@ -5,7 +5,7 @@ A **generated**, zero-dependency typed client for the Bongos API — produced fr
 by hand; it regenerates when the spec changes, so it can never drift from the routes.
 
 - API version: **v1** (served at `/api/bongos/v1`)
-- 355 operations across 55 resource groups
+- 356 operations across 55 resource groups
 
 ## Install
 

package/clients/bongos-client/bongos-client.global.js

@@ -849,6 +849,8 @@ function createClient(opts = {}) {
       deleteTasksIdDependenciesDepId: (args) => request("DELETE", "/tasks/{id}/dependencies/{depId}", { hasBody: false }, args),
       // POST /tasks/{id}/grade — rank: any-builder — POST /tasks/:id/grade
       postTasksIdGrade: (args) => request("POST", "/tasks/{id}/grade", { hasBody: true }, args),
+      // POST /tasks/{id}/merge — rank: metic+archon — POST /tasks/:id/merge
+      postTasksIdMerge: (args) => request("POST", "/tasks/{id}/merge", { hasBody: true }, args),
       // POST /tasks/{id}/override-request — rank: any-builder — POST /tasks/:id/override-request
       postTasksIdOverrideRequest: (args) => request("POST", "/tasks/{id}/override-request", { hasBody: true }, args),
       // POST /tasks/{id}/promote — rank: metic+archon — POST /tasks/:id/promote

package/clients/bongos-client/index.cjs

@@ -848,6 +848,8 @@ function createClient(opts = {}) {
       deleteTasksIdDependenciesDepId: (args) => request("DELETE", "/tasks/{id}/dependencies/{depId}", { hasBody: false }, args),
       // POST /tasks/{id}/grade — rank: any-builder — POST /tasks/:id/grade
       postTasksIdGrade: (args) => request("POST", "/tasks/{id}/grade", { hasBody: true }, args),
+      // POST /tasks/{id}/merge — rank: metic+archon — POST /tasks/:id/merge
+      postTasksIdMerge: (args) => request("POST", "/tasks/{id}/merge", { hasBody: true }, args),
       // POST /tasks/{id}/override-request — rank: any-builder — POST /tasks/:id/override-request
       postTasksIdOverrideRequest: (args) => request("POST", "/tasks/{id}/override-request", { hasBody: true }, args),
       // POST /tasks/{id}/promote — rank: metic+archon — POST /tasks/:id/promote

package/clients/bongos-client/index.d.ts

@@ -371,6 +371,7 @@ export interface PostTasksIdDemoteResponse { ok: boolean }
 export interface PostTasksIdDependenciesRequest { depends_on_task_ids?: unknown[]; depends_on_task_id?: StringifiableId }
 export interface PostTasksIdDependenciesResponse { ok: boolean; added: unknown; dependencies: unknown }
 export interface PostTasksIdGradeRequest { subagent_output?: unknown; signals?: Record<string, unknown>; committed_files?: string[]; diff_hash?: string }
+export interface PostTasksIdMergeRequest { into_task_id: number; reason?: string }
 export interface PostTasksIdOverrideRequestRequest { rationale: string }
 export interface PostTasksIdOverrideRequestResponse { ok: boolean; request: unknown }
 export interface PostTasksIdPublishBranchRequest { branch: string; head_sha: string; bundle_b64: string; value_summary?: string }
@@ -1172,6 +1173,8 @@ export interface BongosClient {
     deleteTasksIdDependenciesDepId(args?: RequestArgs): Promise<DeleteTasksIdDependenciesDepIdResponse>;
     /** POST /tasks/{id}/grade — rank: any-builder */
     postTasksIdGrade(args?: RequestArgs & { body?: PostTasksIdGradeRequest }): Promise<ApiResponse>;
+    /** POST /tasks/{id}/merge — rank: metic+archon */
+    postTasksIdMerge(args: RequestArgs & { body: PostTasksIdMergeRequest }): Promise<ApiResponse>;
     /** POST /tasks/{id}/override-request — rank: any-builder */
     postTasksIdOverrideRequest(args: RequestArgs & { body: PostTasksIdOverrideRequestRequest }): Promise<PostTasksIdOverrideRequestResponse>;
     /** POST /tasks/{id}/promote — rank: metic+archon */

package/clients/bongos-client/index.mjs

@@ -845,6 +845,8 @@ export function createClient(opts = {}) {
       deleteTasksIdDependenciesDepId: (args) => request("DELETE", "/tasks/{id}/dependencies/{depId}", { hasBody: false }, args),
       // POST /tasks/{id}/grade — rank: any-builder — POST /tasks/:id/grade
       postTasksIdGrade: (args) => request("POST", "/tasks/{id}/grade", { hasBody: true }, args),
+      // POST /tasks/{id}/merge — rank: metic+archon — POST /tasks/:id/merge
+      postTasksIdMerge: (args) => request("POST", "/tasks/{id}/merge", { hasBody: true }, args),
       // POST /tasks/{id}/override-request — rank: any-builder — POST /tasks/:id/override-request
       postTasksIdOverrideRequest: (args) => request("POST", "/tasks/{id}/override-request", { hasBody: true }, args),
       // POST /tasks/{id}/promote — rank: metic+archon — POST /tasks/:id/promote