@easysociable/cli 0.1.5 → 0.1.7

package/content-engine/scripts/starter.mjs

@@ -12,6 +12,7 @@ export const ENGINE_PATHS = {
   exports: 'exports',
   topics: 'topics',
   needs: 'needs',
+  hits: 'hits',
   products: 'products',
   recommendations: 'recommendations',
   swipe: 'swipe',

package/content-engine/starter/CLAUDE.md

@@ -42,6 +42,9 @@ This folder is the local **Content Engine** for EasySociable.
 | `profiles/<id>/boundaries.md` | Creator must-avoid (Not covering / Never claim) |
 | `profiles/<id>/expression.md` | Expression likes / dislikes |
 | `profiles/<id>/lanes.md` | Publishing lanes; distinct from pillars |
+| `profiles/<id>/keywords.md` | Search recipes; scan may stage `hits/` |
+| `profiles/<id>/handles.md` | ≤8 ally/foil/competitor handles |
+| `hits/_index.md` | Research triage (not Topics) |
 | `accounts/_index.md` | handle → profile, platforms, voice, overrides |
 | `accounts/overrides/<handle>.md` | Optional account deltas |
 | `platforms/*.md` | Platform craft only |
@@ -72,6 +75,11 @@ This folder is the local **Content Engine** for EasySociable.
 |-------------|--------|
 | capture / 记一下 (+ link) | **§ Capture** only — **not** topics by default |
 | 记需求 / 用户原话 / 评论导出 | **§ Needs** only |
+| 加关键词 / 更新关键词库 | Edit `profiles/<id>/keywords.md` (**ask profile** if unset) |
+| 从需求生成关键词 / 生成 listen\|search\|ask 词 | **§ Keywords · Generate** (chat → confirm → write) |
+| 加/改 handles | Edit `profiles/<id>/handles.md`; keep **≤8** |
+| 扫关键词 / 扫 K-xx / 看 @handle | **§ Keywords** scan (**listen** default) → **§ Hits** |
+| 路由 Hit / H-xx | **§ Hits** → Needs or Capture+library / discard |
 | **扫需求选题** | **§ Topics** from `needs/` where `status=captured`; max 5 `produce` |
 | 扫 inbox 选题 | **§ Topics** — from captures without topic yet (max 5) |
 | 生成 topic (no source specified) | **§ Topics** — offer captured needs or ask which explicit input to use |
@@ -234,8 +242,6 @@ relationships before drafting.
 
 ### Topic generation modes
 
-Use the smallest honest mode:
-
 | mode | Required inputs |
 |------|-----------------|
 | `demand` | Needs + Profile |
@@ -245,8 +251,7 @@ Use the smallest honest mode:
 | `offer_education` | Product + Profile + mechanism |
 | `profile_thesis` | Profile + Pillar |
 
-Needs are the default source. Product or Recommendation presence never proves
-audience demand and never raises a Topic to `ready` by itself.
+**Mode pick (hard):** Before any `produce` Topic write (扫需求选题 / 做成选题 / 生成 topic / 把 N-|C- 做成选题), **ask the user to choose `generation_mode`**. Recommend `demand` when the input is Need-led; do **not** silently assume. Enforce that mode’s required inputs. Product or Recommendation presence never proves audience demand and never raises a Topic to `ready` by itself.
 
 ---
 
@@ -271,6 +276,7 @@ audience demand and never raises a Topic to `ready` by itself.
 - `ready` requires **opp_score ≥ 7.5 and audience ≥ 6**
 - `src_likes` / `src_replies` / `src_bookmarks` (+ views if known) + `engagement_at` on **index** when source is external
 - Do **not** fold `src_*` into opp_score; among **ready** only, prefer replies/bookmarks over views
+- **Language (hard):** Readable judgment + `one_liner` + Decision Card body = profile public language from `voice.md` **Language** (else pillars “Primary public language”, else `engine.json` locale). Verbatim Needs/Capture quotes stay untranslated. Operator chat may add a short gloss in another language; the Topic file and `ready` card body do not. Wrong-language prose → fix before `ready`.
 
 ### Signal paths
 
@@ -290,44 +296,54 @@ audience demand and never raises a Topic to `ready` by itself.
 1. If needs exist with `status=captured`, offer **扫需求选题**.
 2. Otherwise ask which explicit input to use.
 
+0. Resolve profile **Language** (`voice.md` → pillars → `engine.json` locale).
 1. Hard filters (`scoring.md`); else skip.
-2. **Scout gate** — set `purpose` + `scout_label`; if `discard`, do not create T-.
-3. If `produce` (or user forces score):
-   a. Write the **decision card** first (`topics/_template-item.md` upper half): 结论, 状态原因, 适合谁, 核心判断, 下一步, then the readable judgment sections: 要解决的问题, 受众当前相信什么, 我们准备讲哪一刀, 为什么现在, 看完后的改变, 依据, 当前缺口, 建议形式. Keep the summary short; no enum lists, heat, or weighted arithmetic in the readable layer.
+2. **Scout gate** — set `purpose` + `scout_label`; if `discard`, do not create T-. Needs-sourced default `scout_label=demand` (ally/foil/craft_only still allowed when honest).
+3. **Mode pick** — if proceeding toward `produce`, **ask** `generation_mode` (six modes). Recommend `demand` for Need-led input; wait for user choice. Do not write a `produce` T- before the choice. Validate required ids for that mode.
+4. If `produce` (or user forces score):
+   a. Write the **decision card** first (`topics/_template-item.md` upper half), in profile Language: Verdict, Why this status, Who it’s for, Core judgment, Next step, then What problem / What they believe / What cut / Why now / What should change / Based on / Gaps / Suggested form. Keep the summary short; no enum lists, heat, or weighted arithmetic in the readable layer. Set frontmatter `generation_mode` to the chosen mode.
    b. Score all 5 opportunity dims (0–10) → `opp_score`. Put **numbers only** in `## Score`. Do not repeat the card as table notes.
    c. Fill `## Trace` (ids, src heat) and `## Constraints` (writer must-nots). Write `topics/items/T-….md`.
-4. If `library`: optional light item with `status: library`, or capture offer only. Source needs stay `captured` (or `retired` if the user discards).
-5. Add row on `topics/_index.md` (`item` + `src_*` when external). `one_liner` = the card title sentence.
-6. If this T- was `produce` from `N-` ids: update each `needs/_index.md` row + entry to `status=promoted` and `topic_ids` containing this T-. Skip if no T- was written (`discard`).
-7. Output the **Decision Card** in chat — same shape as the readable layer, not a score walkthrough.
+5. If `library`: optional light item with `status: library`, or capture offer only. Source needs stay `captured` (or `retired` if the user discards).
+6. Add row on `topics/_index.md` (`item` + `src_*` when external). `one_liner` = the card’s core-judgment sentence, same language.
+7. If this T- was `produce` from `N-` ids: update each `needs/_index.md` row + entry to `status=promoted` and `topic_ids` containing this T-. Skip if no T- was written (`discard`).
+8. Output the **Decision Card** in chat — same shape and language as the file’s readable layer, not a score walkthrough. Optional operator gloss in another language is chat-only.
 
-**Important:** 「他们现在相信什么」must name the false belief. If you cannot, conflict cannot score above 5. Brief inherits the readable judgment layer, not the Score table.
+**Important:** “What do they believe now?” must name the false belief. If you cannot, conflict cannot score above 5. Brief inherits the readable judgment layer, not the Score table.
 
-### Opportunity Card
+### Decision Card
 
 Paste the decision summary and readable judgment sections. Do not walk five scores in the main body.
 
 ```text
-📋 选题判断卡 T-YYYYMMDD-XX
-【结论】 ready | backlog | parked
-【状态原因】 …
-【适合谁】 …
-【核心判断】 …
-【下一步】 开始生成 / 补需求 / 补证据 / 修改角度 / 放弃
-【要解决的问题】 …
-【他们现在相信什么】 以为 X → 其实 Y
-【我们准备讲哪一刀】 …
-【为什么现在值得讲】 …
-【看完后的改变】 之前 … → 之后 …
-【依据】 …
-【当前缺口】 …
-【建议形式】 …
+📋 Decision card T-YYYYMMDD-XX
+【Verdict】 ready | backlog | parked
+【Why this status】 …
+【Who it’s for】 …
+【Core judgment】 …
+【Next step】 open a run / fill needs / fill evidence / re-angle / kill
+【What problem】 …
+【What they believe】 believe X → actually Y
+【What cut】 …
+【Why now】 …
+【After reading】 before … → after …
+【Based on】 …
+【Gaps】 …
+【Suggested form】 …
 ```
 
 ### Topic → run
 
 On `用 T-xxx 开一单`: **ask platform(s)** if unset (multi-select OK). Create **one run per platform**, same `topic_id`, prefill selection per platform, mark topic `promoted` + list run ids on `topics/_index` `run` column.
 
+### Do not (Topics)
+
+- Create a Topic from a Hit, a scan, or from heat alone
+- Silently pick `generation_mode` (always ask before `produce`)
+- Skip Scout gate or opportunity scoring on `produce` topics
+- Skip the decision card readable layer on `produce`
+- Put `N-` in `capture_id` (needs go in `need_ids` only)
+
 ---
 
 ## § Swipe / atom library
@@ -367,9 +383,46 @@ Hard gate vs first-session (`profile: partial`) vs complete seven dimensions (`p
 
 ---
 
+## § Keywords & handles
+
+`keywords.md` is **three-track**: `listen` (social → Hits), `search` (SEO), `ask` (GEO questions).  
+`handles.md` ≤8 ally/foil/competitor.
+
+**Generate:** from Needs/audience/pillars → chat candidates (max 10) → user confirm → append row with `track` + `seed_from`.  
+**Scan「扫关键词」:** `track=listen` + `active` only (unless user names other ids). Stage **§ Hits**; never Topic; never invent heat/quotes.  
+**search/ask:** not default social scan. Volume/kd optional on search only.
+
+## § Hits (research triage)
+
+**Role:** Sort research hits → **need** | **library** (swipe/atom via Capture+ingest) | **discard**. Not a Topic queue.
+
+| Path | Role |
+|------|------|
+| `hits/_index.md` | Registry |
+| `hits/_template.md` | Entry shape |
+| `hits/entries/H-*.md` | Per-hit files |
+
+Id `H-YYYYMMDD-XX`. Status: `triage` → `routed_need` | `routed_library` | `discarded`.
+
+```text
+🔎 Hit H-…
+【Keyword】 K-xx
+【Source】 @handle · url
+【Kind】 author_post | comment | thread
+【Signal】 demand_language | structure | foil | mixed | noise
+【Heat】 likes … · replies … · bookmarks … · views … @ date
+【Excerpt】 verbatim
+【Suggest】 need | library | discard
+```
+
+On scan: `keyword_id` + `src_*` when known. Route need → **§ Needs**. Route library → **§ Capture** + library offer. Never Hit → Topic.
+
+---
+
 ## § Needs (audience quotes)
 
 Triggers: 记需求 / 用户原话 / 评论导出 / capture audience quotes / user needs.  
+Hits → Needs **only after route confirm** (**§ Hits**).  
 扫需求选题 → **§ Topics** (this section only writes `N-` records).
 
 Follow `easysociable-content/workflows/NEEDS.md` when this engine's copy of the workflow is the newer one; this section wins if both exist.
@@ -439,10 +492,11 @@ User wants a tracked job: "开一单", "open a run", "按完整流程", or a mul
 ### Platform gate (required before folder create)
 
 1. User already named platform(s) → use them.  
-2. Else **ask** (multi-select OK). Offer account’s `platforms` from `accounts/_index` (plus others user names).  
-3. **Do not** silently default to `x` when opening a run (default `x` still applies to other flows, e.g. watchlist filter).  
-4. **N platforms selected → N runs** (same topic/goal/pillar; independent selection/draft/ship).  
-5. Slug: include platform when N>1 or always OK — e.g. `2026-08-26-angle-x`, `…-linkedin`.
+2. Else **stop and ask a required multi-select platform question**. Offer the account’s `platforms` from `accounts/_index` plus any supported platforms the user may name: `x`, `linkedin`, `instagram`, `tiktok`, `threads`, `xiaohongshu`. The user must be able to select one or more.
+3. Do not create a folder, row, brief, or draft before the platform selection is answered.
+4. **Do not** silently default to `x` when opening a run (default `x` still applies to other flows, e.g. watchlist filter).  
+5. **N platforms selected → N runs** (same topic/goal/pillar; independent platform craft, brief, draft, editor, rubric, and ship state).  
+6. Slug: include the platform when N>1 or always OK — e.g. `2026-08-26-angle-x`, `…-linkedin`.
 
 ### Load platform craft (required — every run, every draft stage)
 
@@ -548,7 +602,21 @@ Sharpen these for the specific platform/angle — do not re-derive from scratch.
 
 **Step 3 — Fill brief.md**
 
-Fill Claim (core_thesis + supporting_claims), Evidence (sources, type, quality), Communication (angle, format, hook, structure), Outcome (belief, action, emotion), Constraints, Selection.
+The brief has two layers. The **reading surface** (top) is what a human reviews in 30 seconds. The **appendix** (bottom) is machine handoff data.
+
+Reading surface — fill in this order:
+
+1. **Thesis** — one sentence a stranger can understand.
+2. **For whom → what changes** — reader, their false belief, the after state.
+3. **Evidence** — named sources, type, quality, and what we cannot prove.
+4. **Platform and expression** — platform, form, hook, structure outline.
+5. **Boundaries** — must include / must avoid.
+
+Appendix — fill after the reading surface:
+
+6. **Inherited from Topic** — context, audience, conflict, insight, outcome, evidence (sharpened, not copied).
+7. **Selection** — swipe, atoms, claim ids.
+8. **Quality gate** — 6 × 1–5 scores.
 
 **Step 4 — Brief quality gate (self-scoring)**
 
@@ -624,21 +692,16 @@ Score `rubric.md` (6 × 0–2, max 12).
 【State】 open → next: brief
 ```
 
-**Brief card** (after quality gate passes)
+**Brief card** (after quality gate passes — mirrors brief.md reading surface)
 
 ```text
 📋 Brief  RUN-…
-【Inherited from T-…】
-  Context: …
-  Conflict: current belief → reality
-  Insight: …
-
 【Thesis】 one sentence
-【Evidence】 type · quality (weak/moderate/strong)
-【Angle】 …
-【Outcome】 belief / action / emotion
+【For whom】 reader · false belief → after state
+【Evidence】 sources · type · quality · what we cannot prove
+【Platform】 platform · form · hook · structure outline
+【Boundaries】 must include · must avoid
 【Gate】 Thesis … · Evi … · Angle … · Struct … · Outcome … · Tension … = …/30 ✓
-【Selection】 swipe … · atoms … · claim …
 Reply 简报 OK / 改 …
 ```
 

package/content-engine/starter/engine.json

@@ -7,6 +7,7 @@
     "exports": "exports",
     "topics": "topics",
     "needs": "needs",
+    "hits": "hits",
     "products": "products",
     "recommendations": "recommendations",
     "swipe": "swipe",

package/content-engine/starter/hits/README.md

@@ -0,0 +1,5 @@
+# Hits
+
+Research triage board. Scan → stage here → route to Needs or Library (swipe/atom). Not a Topic queue.
+
+See engine `CLAUDE.md` **§ Hits**.

package/content-engine/starter/hits/_index.md

@@ -0,0 +1,30 @@
+# Hits index (research triage)
+
+**Role:** Staging board for research hits. Route to Needs or Library — **not** a Topic queue.
+
+Newest on top. Scan may write rows here (include source-post heat). Routing out still needs explicit confirm.
+
+| id | date | profile | platform | keyword_id | source | hit_kind | signal | status | excerpt | src_likes | src_replies | src_bookmarks | src_views | engagement_at | file |
+|----|------|---------|----------|------------|--------|----------|--------|--------|---------|-----------|-------------|---------------|-----------|---------------|------|
+
+## Status
+
+| status | Meaning |
+|--------|---------|
+| `triage` | Waiting for route |
+| `routed_need` | Written to `needs/` |
+| `routed_library` | Capture + library ingest path |
+| `discarded` | Rejected |
+
+## Columns
+
+| Column | Meaning |
+|--------|---------|
+| `keyword_id` | Primary recipe `K-xx` |
+| `src_*` | Source-post heat snapshot (not our results) |
+
+## Agent rules
+
+- Exits: **need** \| **library** \| **discard**. Hits are not Topic inputs.
+- On scan: write `keyword_id` + engagement snapshot when known. Do not invent metrics.
+- Author hooks ≠ Needs. Zero starter rows — do not invent hits.

package/content-engine/starter/hits/_template.md

@@ -0,0 +1,46 @@
+---
+id: H-YYYYMMDD-XX
+date: YYYY-MM-DD
+profile: default
+platform: x
+status: triage
+hit_kind: author_post | comment | thread
+signal: demand_language | structure | foil | mixed | noise
+source_keyword:
+keyword_id: K-XX
+source_handle:
+source_url:
+src_likes:
+src_replies:
+src_bookmarks:
+src_views:
+engagement_at:
+---
+
+# H-YYYYMMDD-XX
+
+## Excerpt
+
+> verbatim excerpt — do not rewrite
+
+## Context
+
+Where this came from and why it was kept for triage.
+
+## Engagement snapshot
+
+- likes:
+- replies:
+- bookmarks:
+- views:
+- snapshot_at:
+
+## Suggested route
+
+need | library | discard | undecided
+
+## Route log
+
+| date | action | result_ids |
+|------|--------|------------|
+|  |  |  |

package/content-engine/starter/hits/entries/README.md

@@ -0,0 +1,3 @@
+# Hit entries
+
+One file per `H-YYYYMMDD-XX`. Create from `../_template.md`.

package/content-engine/starter/products/_index.md

@@ -5,6 +5,7 @@ file.
 
 | id | name | status | audience | price | effective_from | file |
 |----|------|--------|----------|-------|----------------|------|
+| P-002 | Content Engine | active | AI-native creators and small content teams building repeatable content workflows | free local install; hosted features not included | 2026-09-16 | products/entries/P-002-content-engine.md |
 | P-001 | EasySociable | active | creators and teams producing branded multi-page social visuals | see entry | 2026-09-04 | products/entries/P-001-easysociable.md |
 
 ## Status values

package/content-engine/starter/products/entries/P-002-content-engine.md

@@ -0,0 +1,136 @@
+---
+id: P-002
+name: Content Engine
+status: active
+effective_from: 2026-09-16
+source_of_truth:
+  - apps/cli/README.md
+  - apps/cli/package.json
+  - apps/cli/content-engine/starter/CLAUDE.md
+  - apps/cli/content-engine/starter/engine.json
+---
+
+# Content Engine
+
+## Product promise
+
+Give AI-native creators and small content teams a local, repeatable workflow
+from audience signals to topics, briefs, drafts, review, feedback, and
+publishable content.
+
+## What it is
+
+Content Engine is a local-first content system installed and operated through
+the `easysociable` command-line tool. It provides a structured local vault,
+profile and audience context, needs, topics, product facts, drafts, runs,
+feedback, and a workbench for inspecting the engine.
+
+It does not require Obsidian. The engine can be initialized and served from a
+workspace directory.
+
+## Best for
+
+- AI-native creators building a repeatable publishing workflow.
+- Agency-of-one operators and small content teams managing recurring content
+  decisions and review.
+- Users who want local files and Agent-assisted workflows instead of a
+  browser-only content writing assistant.
+
+## Not for
+
+- Users looking for guaranteed reach, engagement, conversions, or virality.
+- Users who want an AI model to invent their audience, point of view, proof,
+  or customer results.
+- Users who want a hosted LLM or automatic social-account publishing.
+- Users who only want one prompt or one generated post without a repeatable
+  workflow.
+
+## Problem
+
+AI makes drafting easier, but a blank prompt does not preserve the audience,
+point of view, editorial judgment, review standards, or feedback from the last
+piece. Content Engine turns those decisions into explicit local objects that
+an Agent can use again.
+
+## Mechanism
+
+- Local profile, audience, voice, boundaries, and pillar files provide durable
+  context.
+- Needs and source material become evaluated Topics rather than immediate
+  drafts.
+- Brief, draft, editor, rubric, run, and feedback stages make review and
+  publishing decisions explicit.
+- Product and recommendation libraries keep owned-offer facts separate from
+  audience demand and third-party research.
+- The CLI can initialize, validate, build, serve, and detect local engines.
+
+## Installation
+
+The CLI and local Content Engine are free to install:
+
+```bash
+curl -fsSL https://easysociable.com/install.sh | bash
+easysociable install --skills content,template,slideshow --yes
+easysociable engine init --root /workspace --yes
+```
+
+The installer requires Node.js 20 or newer. The Content Engine workbench also
+requires `python3`. The default CLI skill set currently includes Template and
+Slideshow; the `content` skill is enabled explicitly above.
+
+## Pricing and entitlements
+
+Content Engine and its CLI installation path are free to install. This entry
+does not claim paid hosted features, managed storage, or a hosted LLM.
+
+## Proof
+
+No customer proof, adoption metrics, or performance improvements are recorded.
+
+## Boundaries
+
+- Content Engine owns local content knowledge, editorial workflow, and
+  production tracking.
+- EasySociable is a separate visual production layer for branded templates,
+  slideshow runs, Studio review, and PNG download.
+- The Agent remains responsible for reasoning, content judgment, and supplied
+  evidence.
+- The system does not guarantee that generated content is correct, useful, or
+  successful.
+
+## Claims
+
+### Safe to say
+
+- Install the CLI from the public installer or npm package.
+- Initialize a local Content Engine with `easysociable engine init`.
+- Use local files to preserve profile, audience, needs, topics, briefs, and
+  feedback.
+- Validate and serve the engine through CLI commands.
+
+### Say carefully
+
+- “One-command install” refers to installing the CLI; enabling the Content
+  skill and initializing the engine are additional CLI commands.
+- “Free” refers to the local Content Engine and CLI installation path, not to
+  hosted EasySociable plans or hosted usage.
+- “Gets better” means decisions and feedback persist for the next run; it does
+  not mean the Agent learns autonomously from every review.
+
+### Do not say
+
+- “AI writes and publishes everything automatically.”
+- “The Content Engine guarantees better performance.”
+- “Obsidian is required.”
+- “The hosted EasySociable app is the Content Engine.”
+
+## CTA
+
+- `educate`: explain why content needs durable decisions before generation.
+- `invite`: install the CLI and initialize a local engine.
+- `mention`: show how a Topic, Brief, or review gate survives into the next
+  run.
+
+## Version notes
+
+- 2026-09-16: Added as the local-first Content Engine product.

package/content-engine/starter/profiles/default/README.md

@@ -10,5 +10,8 @@ Replace this starter profile before the first serious draft.
 | `stories.md` | Story inventory (cite or none) |
 | `boundaries.md` | Creator must-avoid (Not covering / Never claim) |
 | `expression.md` | Expression likes / dislikes |
+| `lanes.md` | Publishing lanes |
+| `keywords.md` | Three-track: listen→Hits, search=SEO, ask=GEO |
+| `handles.md` | ≤8 ally/foil/competitor handles |
 
 Accounts: see `accounts/_index.md`.

package/content-engine/starter/profiles/default/handles.md

@@ -0,0 +1,8 @@
+# Handles — profile `default`
+
+**Role:** Small ally / foil / competitor list (≤8). Not a topic radar. Demand language → `./keywords.md`.
+
+| # | platform | handle | niche_tag | why_on_list | weight | status |
+|---|----------|--------|-----------|-------------|--------|--------|
+
+Zero starter rows. Add only with a concrete ally/foil/competitor reason.

package/content-engine/starter/profiles/default/keywords.md

@@ -0,0 +1,31 @@
+# Keywords — profile `default`
+
+**Role:** Three-track recipes — **listen** (social), **search** (SEO), **ask** (GEO). Not topics.
+
+| track | Means | Default use |
+|-------|--------|-------------|
+| `listen` | Social search strings | 「扫关键词」→ Hits |
+| `search` | Google/SEO phrases | Content map; volume optional; not default social scan |
+| `ask` | 15–25 word AI-engine questions | GEO coverage; not social scan |
+
+**Scan rule:** 「扫关键词」→ `track=listen` + `active` only → Hits. Never scan → Topic.
+
+| id | track | query | platform | intent | layer | pillar | why | weight | status | seed_from | pair_id | volume | kd | serp_intent | volume_source | volume_at | ask_engines | last_scanned |
+|----|-------|-------|----------|--------|-------|--------|-----|--------|--------|-----------|---------|--------|----|-------------|---------------|-----------|-------------|--------------|
+| K-01 | listen | (fill audience pain phrase) | x | pain | L1 | P1 | (why) | 1.0 | paused | manual |  |  |  |  |  |  |  |  |
+
+Replace fill rows after interview / first needs. Prefer 8–12 active **listen** seeds before routine scans.
+
+## Column notes
+
+| Column | Meaning |
+|--------|---------|
+| `track` | `listen` \| `search` \| `ask` |
+| `seed_from` | `need:N-…` / `hit:H-…` / `audience` / `pillar:P…` / `manual` |
+| `volume` / `kd` | search annotations; empty OK |
+
+## Agent rules
+
+- Default scan: listen + active only; max 5
+- Propose new rows in chat; write after confirm
+- Do not invent Need quotes from keywords

package/content-engine/starter/runs/_template/brief.md

@@ -6,52 +6,62 @@ gate_pass: false
 
 # Brief
 
-## Inherited (from Topic T-xxx — do not re-derive; sharpen only)
+## Thesis
 
-- **context:**
-- **audience:**
-- **conflict:**
-- **insight:**
+One sentence a stranger can understand without any product context.
 
-## Claim
+(fill)
 
-- **core_thesis:**
-- **supporting_claims:**
-  1.
-  2.
-  3.
+## For whom → what changes
+
+**Reader:** (who specifically has this problem)
+
+**They currently believe:** (the false belief or stuck assumption)
+
+**After reading:** (what they should believe, do, or stop doing)
 
 ## Evidence
 
-- **sources:**
-- **type:** data | example | case | mechanism | expert | observation
-- **quality:** weak | moderate | strong
+**What we have:** (named sources — Need ids, product facts, mechanisms, cases)
+
+**Evidence type:** data | example | case | mechanism | expert | observation
+
+**Quality:** weak | moderate | strong
 
-## Communication
+**What we cannot prove:** (be honest — no invented outcomes)
 
-- **angle:**
-- **format:**
-- **hook:**
-- **structure:**
+## Platform and expression
 
-## Outcome
+- **Platform:** x | linkedin | ig | tiktok
+- **Form:** short post | thread | carousel | …
+- **Hook:** (first line or opening move)
+- **Structure:** (the argument flow in one line, e.g. hook → false belief → mechanism → action)
+
+## Boundaries
+
+- **Must include:** (non-negotiable claims or framing)
+- **Must avoid:** (hard bans for this draft; merge `profiles/<id>/boundaries.md` + voice Do not say)
+
+---
 
-- **belief:**
-- **action:**
-- **emotion:**
+## Appendix (machine / handoff — not the reading surface)
 
-## Constraints
+### Inherited from Topic
 
-- **must_include:**
-- **must_avoid:** (merge `profiles/<id>/boundaries.md` + voice Do not say)
+- context:
+- audience:
+- conflict:
+- insight:
+- outcome:
+- evidence:
 
-## Selection (required; none allowed)
+### Selection
 
-- **swipe:** id | none — why:
-- **atoms:** hook:… | cta:… | proof:… | none — why:
-- **claim:** id | none — why:
+- swipe: id | none — why
+- atoms: hook:… | cta:… | none — why
+- claim: id | none — why
 
-## Brief quality gate
+### Quality gate
 
 | # | Gate | Score 1–5 | Note |
 |---|------|-----------|------|