From a25df3c3bc876ec20b94654111713bf2ae8e0d72 Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:32:54 +0800 Subject: [PATCH 01/11] feat(expression): put a research-depth upper witness on freeform path (#836) --- .../references/freeform-answers.md | 30 +++++++++++-------- 1 file changed, 18 insertions(+), 12 deletions(-) diff --git a/skills/fomo-kernel/references/freeform-answers.md b/skills/fomo-kernel/references/freeform-answers.md index e7892cf..8a43549 100644 --- a/skills/fomo-kernel/references/freeform-answers.md +++ b/skills/fomo-kernel/references/freeform-answers.md @@ -1,24 +1,30 @@ # Freeform informational answers -## The exemplar +## The exemplars -One whole freeform answer. This route adds no parameter of its own (rule 4 -below), so the exemplar is what this file has to show: the one position the -question turns on, the rest of the computed inventory behind a single offer, -and one end block. Read it before writing one. +Two ends of one rule live on this surface. A simple lookup should still look +like the compact `freeform_cash_question` or `freeform_positions_view` scenes +in the corpus; do not inflate it. The fenced scene below is the **upper +witness**: when a live decision genuinely turns on several sourced increments, +the increment gate earns several blocks. It is not a minimum length, a default +answer size, or permission to narrate research that does not change the call. It is copied verbatim from `tests/agent/expression-witnesses.json`, scene -`freeform_positions_view` — the corpus that +`freeform_research_depth` — the corpus that §3.5 of the repository's expression contract (`docs/expression-contract.md`) makes the -binding statement of this shape. Every issuer in it is invented, and -`tests/test_expression_contract.py` fails if the two copies disagree. +binding statement of this shape. Every issuer and source in it is invented, +and `tests/test_expression_contract.py` fails if the two copies disagree. -```exemplar freeform_positions_view -六檔裡只有一檔值得你現在看：GRDC 佔 31.4%，其餘五檔全在 10% 以下。 +```exemplar freeform_research_depth +先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。 -最大的那筆已經是第二大的三倍，其他五檔加起來還不到它。要完整的逐檔表（股數、成本、市值、損益、診斷標籤）跟我說一聲就給。 +交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。 -（帳本 8/14；價格 8/14 收盤；ETF 未拆解，成分重疊未評） +形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。 + +會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，要增加曝險，先把部位降回 30% 內。 + +（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值） ``` The user does not only meet this product through `prepare → preview → From 75695955481cf937a5a0caf31fdc7e0bcd1b58a2 Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:36:08 +0800 Subject: [PATCH 02/11] test(expression): add freeform research-depth upper witness (#836) --- tests/agent/expression-witnesses.json | 53 ++++++++++++++++++++++++++- 1 file changed, 52 insertions(+), 1 deletion(-) diff --git a/tests/agent/expression-witnesses.json b/tests/agent/expression-witnesses.json index ffb5ccf..69af046 100644 --- a/tests/agent/expression-witnesses.json +++ b/tests/agent/expression-witnesses.json @@ -1,7 +1,7 @@ { "schema_version": 2, "privacy": "synthetic_only", - "comment": "Exemplars for docs/expression-contract.md. #832 made this corpus the binding statement of the answer pyramid (§3): three to five canonical exemplars per conversational surface, each declaring the one-sentence answer it leads with (`core`) and the increment each of its blocks adds (`blocks[].adds`), plus counter-exemplars for the four named bans. #825 retired the formatting-only E-1 through E-4; placement, markers and answer length still vary freely. #830 added E-6 and the owner's three approved acceptance templates. Every scene is invented, nothing is derived from a user record, and all issuers are fictional (Widgetron WDGT, Gridcore GRDC, Fabrion FABR, ACME).", + "comment": "Exemplars for docs/expression-contract.md. #832 made this corpus the binding statement of the answer pyramid (§3): three to five canonical exemplars per conversational surface, each declaring the one-sentence answer it leads with (`core`) and the increment each of its blocks adds (`blocks[].adds`), plus counter-exemplars for the four named bans. #825 retired the formatting-only E-1 through E-4; placement, markers and answer length still vary freely. #830 added E-6 and the owner's three approved acceptance templates. #836 adds a research-depth upper witness without reinstating a length cap. Every scene is invented, nothing is derived from a user record, and all issuers and sources are fictional (Widgetron WDGT, Gridcore GRDC, Fabrion FABR, ACME, Northstar Ratings).", "surfaces": { "consider": "The deterministic portfolio-consequence route (references/trade-consequence.md).", "no_book": "A live decision with no recorded book (references/decision-framing.md).", @@ -347,6 +347,57 @@ ], "answer": "六檔裡只有一檔值得你現在看：GRDC 佔 31.4%，其餘五檔全在 10% 以下。\n\n最大的那筆已經是第二大的三倍，其他五檔加起來還不到它。要完整的逐檔表（股數、成本、市值、損益、診斷標籤）跟我說一聲就給。\n\n（帳本 8/14；價格 8/14 收盤；ETF 未拆解，成分重疊未評）" }, + { + "id": "freeform_research_depth", + "surface": "freeform", + "kind": "positive", + "covers": [ + "D1", + "D2", + "D4", + "D5", + "D7", + "C1", + "C2" + ], + "note": "The upper witness #836 was missing: a freeform decision whose length is earned by several distinct, sourced increments. Compact freeform scenes remain in the corpus; this one proves that the increment gate, not exemplar character count, decides when research depth belongs in the answer.", + "core": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。", + "blocks": [ + { + "text": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。", + "adds": "the stance, the user-rule collision, and the unresolved deal premise that decide it" + }, + { + "text": "交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管", + "adds": "the financing structure and the regulatory contingency" + }, + { + "text": "新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後", + "adds": "the timing mismatch between capacity gained and earnings delivered" + }, + { + "text": "形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察", + "adds": "the balance-sheet consequence the headline alone does not show" + }, + { + "text": "這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻", + "adds": "why no-add is not the same decision as sell" + }, + { + "text": "會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑", + "adds": "the two observable falsifiers that would reopen the add" + }, + { + "text": "在那之前，要增加曝險，先把部位降回 30% 內", + "adds": "the smallest action that makes another add compatible with the user's rule" + }, + { + "text": "（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）", + "adds": "the caliber block: book, pricing session, outside sources, and material gaps" + } + ], + "answer": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。\n\n會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，要增加曝險，先把部位降回 30% 內。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" + }, { "id": "freeform_degraded_pricing", "surface": "freeform", From fdd387597ce1e2c0b30b341d0abea32d38c36ef1 Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:42:25 +0800 Subject: [PATCH 03/11] fix(expression): make the upper-witness action consistent with its weights (#836) --- skills/fomo-kernel/references/freeform-answers.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/skills/fomo-kernel/references/freeform-answers.md b/skills/fomo-kernel/references/freeform-answers.md index 8a43549..5c33b92 100644 --- a/skills/fomo-kernel/references/freeform-answers.md +++ b/skills/fomo-kernel/references/freeform-answers.md @@ -22,7 +22,7 @@ and `tests/test_expression_contract.py` fails if the two copies disagree. 形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。 -會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，要增加曝險，先把部位降回 30% 內。 +會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。 （帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值） ``` From 49d15bbfbe370f4e3b0cd2f6790014a87e0699ff Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:45:27 +0800 Subject: [PATCH 04/11] test(expression): align upper-witness action with pre-trade weight (#836) --- tests/agent/expression-witnesses.json | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tests/agent/expression-witnesses.json b/tests/agent/expression-witnesses.json index 69af046..b8d61ec 100644 --- a/tests/agent/expression-witnesses.json +++ b/tests/agent/expression-witnesses.json @@ -164,7 +164,7 @@ }, { "text": "我看不到你在別處的餘額", - "adds": "the boundary: balances outside the recorded book, stated once" + "adds": "the boundary: balances outside the book, stated once" }, { "text": "最小的一刀是減掉現在最大的那筆的三分之一", @@ -388,7 +388,7 @@ "adds": "the two observable falsifiers that would reopen the add" }, { - "text": "在那之前，要增加曝險，先把部位降回 30% 內", + "text": "在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己", "adds": "the smallest action that makes another add compatible with the user's rule" }, { @@ -396,7 +396,7 @@ "adds": "the caliber block: book, pricing session, outside sources, and material gaps" } ], - "answer": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。\n\n會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，要增加曝險，先把部位降回 30% 內。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" + "answer": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。\n\n會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" }, { "id": "freeform_degraded_pricing", From ac359510621e3d1b10d46763554c6e8c9e206f7e Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:49:08 +0800 Subject: [PATCH 05/11] chore(expression): keep unrelated witness wording unchanged (#836) --- tests/agent/expression-witnesses.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/agent/expression-witnesses.json b/tests/agent/expression-witnesses.json index b8d61ec..82058f4 100644 --- a/tests/agent/expression-witnesses.json +++ b/tests/agent/expression-witnesses.json @@ -164,7 +164,7 @@ }, { "text": "我看不到你在別處的餘額", - "adds": "the boundary: balances outside the book, stated once" + "adds": "the boundary: balances outside the recorded book, stated once" }, { "text": "最小的一刀是減掉現在最大的那筆的三分之一", From f6d543662f9d2450cd8215833395347b75d760ed Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:51:32 +0800 Subject: [PATCH 06/11] fix(expression): keep the user cap outside the research falsifier (#836) --- skills/fomo-kernel/references/freeform-answers.md | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/skills/fomo-kernel/references/freeform-answers.md b/skills/fomo-kernel/references/freeform-answers.md index 5c33b92..2f8c09c 100644 --- a/skills/fomo-kernel/references/freeform-answers.md +++ b/skills/fomo-kernel/references/freeform-answers.md @@ -22,7 +22,7 @@ and `tests/test_expression_contract.py` fails if the two copies disagree. 形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。 -會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。 +會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。 （帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值） ``` @@ -166,8 +166,7 @@ re-deriving any of it, the same discipline this file already applies to every other engine number. Read its `sizing` block too, and relay what it says (#737). It carries the -engine's own verdict on the weights it just emitted: whether a weight could -be computed at all, the `aggregate_currency` those weights are measured in, +engine's own verdict on the weights it just emitted: whether a weight could be computed at all, the `aggregate_currency` those weights are measured in, and — when one could not — every holding that has none, each with the engine's own reason, plus the missing prices or FX rates that explain it. Two things follow. A mixed-currency book's `value` is stated in each From c96b8b20beaffe9ad44567b6c291a92cebd0e293 Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:54:12 +0800 Subject: [PATCH 07/11] test(expression): preserve the user cap across research reopening (#836) --- tests/agent/expression-witnesses.json | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tests/agent/expression-witnesses.json b/tests/agent/expression-witnesses.json index 82058f4..417a09d 100644 --- a/tests/agent/expression-witnesses.json +++ b/tests/agent/expression-witnesses.json @@ -384,8 +384,8 @@ "adds": "why no-add is not the same decision as sell" }, { - "text": "會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑", - "adds": "the two observable falsifiers that would reopen the add" + "text": "會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑", + "adds": "the two observable signals that would reopen an in-rule add" }, { "text": "在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己", @@ -396,7 +396,7 @@ "adds": "the caliber block: book, pricing session, outside sources, and material gaps" } ], - "answer": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。\n\n會讓我改口的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" + "answer": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。\n\n會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" }, { "id": "freeform_degraded_pricing", From ab9aca2a3d79599f9b763c5d1dc97c3b6789814e Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 14:59:07 +0800 Subject: [PATCH 08/11] chore(expression): remove unrelated line-wrap diff (#836) --- skills/fomo-kernel/references/freeform-answers.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/skills/fomo-kernel/references/freeform-answers.md b/skills/fomo-kernel/references/freeform-answers.md index 2f8c09c..6dd6b67 100644 --- a/skills/fomo-kernel/references/freeform-answers.md +++ b/skills/fomo-kernel/references/freeform-answers.md @@ -166,7 +166,8 @@ re-deriving any of it, the same discipline this file already applies to every other engine number. Read its `sizing` block too, and relay what it says (#737). It carries the -engine's own verdict on the weights it just emitted: whether a weight could be computed at all, the `aggregate_currency` those weights are measured in, +engine's own verdict on the weights it just emitted: whether a weight could +be computed at all, the `aggregate_currency` those weights are measured in, and — when one could not — every holding that has none, each with the engine's own reason, plus the missing prices or FX rates that explain it. Two things follow. A mixed-currency book's `value` is stated in each From f9e7b99ee1a79299fa72826dfd5438046bfe6d15 Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 15:01:02 +0800 Subject: [PATCH 09/11] fix(expression): align the upper-witness stance with its smaller alternative (#836) --- skills/fomo-kernel/references/freeform-answers.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/skills/fomo-kernel/references/freeform-answers.md b/skills/fomo-kernel/references/freeform-answers.md index 6dd6b67..a6b2ca9 100644 --- a/skills/fomo-kernel/references/freeform-answers.md +++ b/skills/fomo-kernel/references/freeform-answers.md @@ -16,11 +16,11 @@ binding statement of this shape. Every issuer and source in it is invented, and `tests/test_expression_contract.py` fails if the two copies disagree. ```exemplar freeform_research_depth -先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。 +不要照原尺寸加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。 交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。 -形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。 +形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把部位推過上限，等於同時押融資、監管與整合三件事。 會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。 From 69194e006a896184bbdfc7842570642284beb4ab Mon Sep 17 00:00:00 2001 From: TING Date: Mon, 24 Aug 2026 15:03:46 +0800 Subject: [PATCH 10/11] test(expression): align the upper-witness stance and current state (#836) --- tests/agent/expression-witnesses.json | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/tests/agent/expression-witnesses.json b/tests/agent/expression-witnesses.json index 417a09d..5f362a8 100644 --- a/tests/agent/expression-witnesses.json +++ b/tests/agent/expression-witnesses.json @@ -151,7 +151,7 @@ "D7", "C1" ], - "note": "#778's acceptance shape. A negative post-trade cash balance is owed as two numbers, and the decision those numbers imply — this needs funding or a sale — is what the top floor carries. The boundary that the engine cannot see accounts outside the recorded book is still stated, once, beside the claim it qualifies, rather than becoming the lead.", + "note": "#778's acceptance shape. A negative post-trade cash balance is owed as two numbers, and the decision they imply — this needs funding or a sale — is what the top floor carries. The boundary that the engine cannot see accounts outside the recorded book is still stated, once, beside the claim it qualifies, rather than becoming the lead.", "core": "這筆要成立，你得先賣掉一些東西，或者動用這本帳本看不到的現金", "blocks": [ { @@ -361,10 +361,10 @@ "C2" ], "note": "The upper witness #836 was missing: a freeform decision whose length is earned by several distinct, sourced increments. Compact freeform scenes remain in the corpus; this one proves that the increment gate, not exemplar character count, decides when research depth belongs in the answer.", - "core": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。", + "core": "不要照原尺寸加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。", "blocks": [ { - "text": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。", + "text": "不要照原尺寸加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。", "adds": "the stance, the user-rule collision, and the unresolved deal premise that decide it" }, { @@ -396,7 +396,7 @@ "adds": "the caliber block: book, pricing session, outside sources, and material gaps" } ], - "answer": "先不要加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把超過上限的部位繼續放大，等於同時押融資、監管與整合三件事。\n\n會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" + "answer": "不要照原尺寸加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把部位推過上限，等於同時押融資、監管與整合三件事。\n\n會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" }, { "id": "freeform_degraded_pricing", From ff8f11c6030898f2b22a5f541e3a35d220a405ad Mon Sep 17 00:00:00 2001 From: TING Date: Thu, 27 Aug 2026 16:12:11 +0800 Subject: [PATCH 11/11] fix(expression): keep provenance on one floor (#836) --- skills/fomo-kernel/references/freeform-answers.md | 2 +- tests/agent/expression-witnesses.json | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/skills/fomo-kernel/references/freeform-answers.md b/skills/fomo-kernel/references/freeform-answers.md index a6b2ca9..d49e82d 100644 --- a/skills/fomo-kernel/references/freeform-answers.md +++ b/skills/fomo-kernel/references/freeform-answers.md @@ -24,7 +24,7 @@ and `tests/test_expression_contract.py` fails if the two copies disagree. 會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。 -（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值） +（帳本 8/14、價格 8/14 收盤；未評反壟斷通過機率與交易後估值） ``` The user does not only meet this product through `prepare → preview → diff --git a/tests/agent/expression-witnesses.json b/tests/agent/expression-witnesses.json index 5f362a8..dc40ef1 100644 --- a/tests/agent/expression-witnesses.json +++ b/tests/agent/expression-witnesses.json @@ -151,7 +151,7 @@ "D7", "C1" ], - "note": "#778's acceptance shape. A negative post-trade cash balance is owed as two numbers, and the decision they imply — this needs funding or a sale — is what the top floor carries. The boundary that the engine cannot see accounts outside the recorded book is still stated, once, beside the claim it qualifies, rather than becoming the lead.", + "note": "#778's acceptance shape. A negative post-trade cash balance is owed as two numbers, and the decision those numbers imply — this needs funding or a sale — is what the top floor carries. The boundary that the engine cannot see accounts outside the recorded book is still stated, once, beside the claim it qualifies, rather than becoming the lead.", "core": "這筆要成立，你得先賣掉一些東西，或者動用這本帳本看不到的現金", "blocks": [ { @@ -392,11 +392,11 @@ "adds": "the smallest action that makes another add compatible with the user's rule" }, { - "text": "（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）", - "adds": "the caliber block: book, pricing session, outside sources, and material gaps" + "text": "（帳本 8/14、價格 8/14 收盤；未評反壟斷通過機率與交易後估值）", + "adds": "the caliber block: book, pricing session, and material gaps" } ], - "answer": "不要照原尺寸加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把部位推過上限，等於同時押融資、監管與整合三件事。\n\n會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；來源：WDGT 8-K／投資人簡報 8/12、Northstar Ratings 8/13；未評反壟斷通過機率與交易後估值）" + "answer": "不要照原尺寸加 WDGT：這筆會把部位從 28.6% 推到 32.4%，先跨過你自己寫的 30% 上限，而收購的價值仍取決於尚未驗證的融資與整合。\n\n交易不是免費增長：WDGT 為 Fabrion 報價 $4.2B，預計 60% 舉債、40% 發股，交割還要過監管（WDGT 8-K，2026-08-12）。新增的是 18% 產能，交割時只帶來 6% EBITDA；管理層把 70% 協同效益放在十八個月後（WDGT 投資人簡報，2026-08-12），所以營運好處和資產負債表壓力不是同時兌現。\n\n形式上淨負債／EBITDA 會從 2.1 倍升到 3.9 倍，Northstar Ratings 已列入負向觀察（2026-08-13）。這不是賣出理由：現有部位的核心 thesis 還沒有被公告本身推翻；但在證據到位前，把部位推過上限，等於同時押融資、監管與整合三件事。\n\n會讓我重新評估 30% 內小幅加碼的證據有兩個：最終融資把舉債比例壓到 40% 以下，或交割後兩季的協同效益 run-rate 達到管理層路徑。在那之前，若仍要增加曝險，就把這筆縮到買後仍不超過 30%；不要先跨過自己的線，再等交易替你證明自己。\n\n（帳本 8/14、價格 8/14 收盤；未評反壟斷通過機率與交易後估值）" }, { "id": "freeform_degraded_pricing",