@easysociable/cli 0.1.1 → 0.1.3

package/content-engine/starter/topics/_template-item.md

@@ -0,0 +1,90 @@
+---
+id: T-YYYYMMDD-XX
+date: YYYY-MM-DD
+status: scored
+account: your-handle
+profile: default
+platform: x
+pillar: P1
+source_type: watchlist | inbox | user | discovery
+purpose: produce | library | discard
+scout_label: demand | ally | foil | craft_only
+---
+
+# Opportunity
+
+One sentence — what's the content opportunity, not a title.
+
+## Source
+
+- **type:** watchlist | inbox | user | discovery
+- **url:**
+- **handle:**
+- **discovery_id:** (only if promoted from D-…; else empty)
+- **capture_id:** (only if linked to inbox; else empty)
+- **raw:** (only if exists)
+- **engagement snapshot:** likes / replies / bookmarks / views @ time
+
+## Scout
+
+- **purpose:** produce | library
+- **scout_label:** demand | ally | foil | craft_only
+- **scout_note:** (one line — why not discard)
+
+## Context
+
+- **what_changed:**
+- **why_now:**
+
+## Audience
+
+- **who:**
+- **current_belief:**
+- **icp_checks:** (pain / buyer shape / wedge — note which hit, from profile)
+
+## Conflict
+
+- **old_model:**
+- **new_reality:**
+- **belief_gap:**
+
+## Insight
+
+- **potential_insight:**
+- **why_nonobvious:**
+
+## Evidence
+
+- **available:**
+- **type:** data | example | case | mechanism | expert | observation
+- **gaps:**
+
+## Score (opportunity)
+
+| dim | score 0-10 | note |
+|-----|------------|------|
+| context | | |
+| audience | | |
+| conflict | | |
+| insight | | |
+| evidence | | |
+
+**opp_score:** (weighted per `topics/scoring.md`)  
+**audience_floor:** (hard check ≥ 6)
+
+## Profile fit
+
+- pillars:
+- audience_pain:
+- risk:
+
+## Suggested production (optional, not binding)
+
+- swipe: none | S…
+- atoms: none | …
+- claim: none | C…
+
+## Decision
+
+- next: open_run | hold | kill | library_ingest | research_evidence
+- run_id: (filled when promoted)

package/content-engine/starter/topics/items/README.md

@@ -0,0 +1,4 @@
+# items/
+
+One markdown file per topic (`T-YYYYMMDD-XX.md`).  
+Registry: `../_index.md` (column `item` points here).

package/content-engine/starter/topics/scoring.md

@@ -0,0 +1,201 @@
+# Topic scoring — Opportunity evaluation
+
+Stable rules for evaluating **Content Opportunities** for the active profile.  
+Scores whether a topic **deserves to exist as content**, not whether it can be written well.
+
+**Principle:** Don't generate content from topics. Generate content from validated opportunities.
+
+**Goal of a `ready` opportunity:** worth opening a run that speaks to the readers and pains in  
+`profiles/<id>/audience.md` and `pillars.md`, and can land the account’s product/wedge honestly  
+(not empty consensus, not hard sell).
+
+---
+
+## Pipeline (do not skip)
+
+```text
+watchlist / inbox / url / discovery signal
+  → 0) Hard filters
+  → 1) Scout gate (purpose + label)     ← most signals die here
+  → 2) Opportunity five-dim score        ← only for produce candidates
+  → 3) Verdict (opp_score + audience floor)
+  → 4) Among ready only: sort by src_*
+```
+
+Library-only signals may skip (2)–(3) — see Scout purposes below.
+
+Watchlist scans create **Discovery** records first. Topic scoring runs only after the user explicitly promotes a Discovery (or names another explicit input).
+
+---
+
+## 0) Hard filters (fail → do not create topic, or `parked`)
+
+- Off **profile pillars** (unless user override)
+- Hits audience **Not for** (F4F, pure crypto shill, comment-gate junk as the only substance, etc.)
+- Requires invented personal metrics/experience to write honestly
+- Wrong network for the watchlist row (e.g. X tools on a `tiktok` handle)
+
+---
+
+## 1) Scout gate (before opportunity scores)
+
+Every scanned signal gets a **purpose** and a **scout_label**.  
+**Default bias: discard.** Max **5** `produce` topics per promotion session.
+
+### Purpose
+
+| purpose | Meaning | Create T-? | Score opportunity? |
+|---------|---------|------------|------------------|
+| `discard` | Noise / wrong ICP / no reusable mechanism | **No** | No |
+| `library` | Worth swipe/atom/claim later; not this week's post | Optional T- with status `library` **or** offer capture only | Optional; **not** `ready` |
+| `produce` | Candidate to open a run | **Yes** | **Yes** |
+
+### Scout label (on produce / library)
+
+| scout_label | Meaning |
+|-------------|---------|
+| `demand` | Target reader states a real pain/job in their own words |
+| `ally` | Peer system / adjacent thinking we can teach beside |
+| `foil` | Competitor or default path to contrast |
+| `craft_only` | Stealable structure; weak ICP/wedge fit — prefer `library`, rarely `produce` |
+
+### Scout discard examples (profile-aware)
+
+- Pure growth/engagement playbook with **zero** angle toward this profile’s pillars
+- Comment-gate kits with no inspectable mechanism
+- End-user vanity metrics with no buyer/operator shape from `audience.md`
+- Wrong vertical vs profile **Not for** (unless user override)
+
+### Scout pass hints (toward `produce`)
+
+- Pain named in `audience.md` (or clear operator inside that buyer shape)
+- Mechanism / contrast that maps to at least one pillar
+- Ally or foil that sharpens the account’s wedge without inventing claims
+- Reusable structure that still clears the audience floor (else prefer `library`)
+
+Record on the topic item: `purpose`, `scout_label`, one-line `scout_note`.
+
+---
+
+## 2) Opportunity dimensions (0–10 each)
+
+Only required for `purpose: produce`. Optional brief notes for `library`.
+
+**Principle:** these five dimensions answer **"is this worth saying?"** — not "can it be written well?"  
+Presentation concerns (framework, structure, format) belong at Brief stage.
+
+| Key | Name | What to judge | Weight |
+|-----|------|---------------|--------|
+| `context` | Context | What changed? Why now? Timing, freshness, relevance to the audience's world | **15%** |
+| `audience` | Audience | Who specifically cares? ICP match — pain / buyer shape / wedge from profile | **20%** |
+| `conflict` | Conflict | Current belief vs reality — is there a real gap? How strong is the tension? No conflict = no content | **25%** |
+| `insight` | Insight | Non-obvious perspective, mechanism, or reframe the audience hasn't seen | **25%** |
+| `evidence` | Evidence | Can the core angle be supported with something specific? What evidence exists or could be found? | **15%** |
+
+```text
+opp_score = 0.15×context + 0.20×audience + 0.25×conflict + 0.25×insight + 0.15×evidence
+```
+
+Round to 1 decimal.
+
+### Per-dimension scoring guide
+
+**Context (15%)**
+
+| Range | Guide |
+|-------|-------|
+| 8–10 | Clear, recent shift the audience has felt but may not have named |
+| 5–7 | Real change but not urgent or widely felt yet |
+| 0–4 | Nothing changed; evergreen truism or stale news |
+
+**Audience (20%)** — read `profiles/<id>/audience.md` and `pillars.md` first
+
+Score using these checks (note which hit):
+
+1. **Pain** — matches a real pain / job from audience (or clear operator inside that shape)
+2. **Buyer shape** — who buys/acts matches audience buyer as defined there
+3. **Wedge land** — can honestly teach or contrast toward this account’s product/wedge without inventing claims
+
+| Range | Guide |
+|-------|-------|
+| 8–10 | ≥2 checks strong; clear buyer + pain |
+| 6–7 | One strong check; usable with careful angle |
+| 4–5 | Stretch; usually `backlog` even if other dims are high |
+| 0–3 | Wrong room → should have been scout `discard` / `library` |
+
+**Conflict (25%)** — the most important dimension
+
+| Range | Guide |
+|-------|-------|
+| 8–10 | Clear belief gap: audience believes X, reality is Y, and the gap has consequences |
+| 5–7 | Some tension but the gap is soft or already widely acknowledged |
+| 0–4 | No real conflict; topic is an observation or list with nothing to challenge |
+
+When scoring, the agent must write the `current_belief` and `belief_gap` on the topic item — not just a score. If you can't articulate what the audience currently believes wrong, the conflict score cannot be above 5.
+
+**Insight (25%)**
+
+| Range | Guide |
+|-------|-------|
+| 8–10 | Genuinely non-obvious; reframes how the audience thinks about the problem |
+| 5–7 | Useful perspective but predictable for a knowledgeable reader |
+| 0–4 | Restates common knowledge |
+
+**Evidence (15%)**
+
+| Range | Guide |
+|-------|-------|
+| 8–10 | Specific, named evidence available (data, product behavior, case, mechanism) |
+| 5–7 | Evidence exists but generic or secondhand |
+| 0–4 | No evidence beyond assertion; "experts say" with no expert named |
+
+When `evidence < 5`, agent should flag an **evidence gap** on the topic item and may suggest **research** before finalizing the score. Topics can be re-scored after research fills the gap.
+
+---
+
+## 3) Verdict
+
+For `purpose: produce`:
+
+| Verdict | Rule |
+|---------|------|
+| **ready** | `opp_score ≥ 7.5` **and** `audience ≥ 6` |
+| **backlog** | opp_score ≥ 6.0 and audience ≥ 5, but not both ready thresholds; **or** one dim very high / another mid — may research, re-angle, and rescore |
+| **parked** | opp_score < 6.0 **or** audience < 5 **or** hard filter |
+
+For `purpose: library`: status `library` (not ready). User may later promote via re-angle + rescore as `produce`.
+
+**Never** promote to `ready` on high `src_*` alone.
+
+### Weak-dim guidance
+
+When any single dim scores < 6, agent flags it with a suggestion:
+
+| Weak dim | Suggestion |
+|----------|-----------|
+| context | "Is there a real trigger or change? If not, consider timing or reframing." |
+| audience | "Who specifically has this problem? If no ICP match, consider library instead." |
+| conflict | "What does the audience believe that's wrong? If no gap, this may not be content." |
+| insight | "What's the non-obvious angle? If predictable, consider a different perspective." |
+| evidence | "Can we prove this? Research X to strengthen." |
+
+---
+
+## 4) Source engagement (sort only)
+
+When source is an external post, record on **`topics/_index.md`**:
+
+- `src_likes`, `src_replies`, `src_bookmarks`, optional `src_views`
+- `engagement_at`
+
+**Do not** fold into `opp_score`.  
+Among **ready** items only: prefer higher `src_replies` / `src_bookmarks` over raw likes/views.  
+High views + low audience → still not ready.
+
+---
+
+## Rescore
+
+Allowed: same `T-id`, update scores / purpose / status in item + `_index`.  
+Re-angling a `library` or `backlog` into profile fit is the usual upgrade path.  
+Research → new evidence → re-score `evidence` dim is the other upgrade path.

package/content-engine/starter/topics/watchlists/README.md

@@ -0,0 +1,10 @@
+# topics/watchlists/
+
+**One file per profile** + optional shared table.
+
+| Path | Scope |
+|------|--------|
+| `_shared.md` | All profiles (`*`) — merged on every scan |
+| `<profile_id>.md` | Only that profile (e.g. `easysociable-builder.md`) |
+
+Registry / scan rules: `../WATCHLIST.md`.

package/content-engine/starter/topics/watchlists/_shared.md

@@ -0,0 +1,14 @@
+# Watchlist — shared (`*`)
+
+Rows here apply to **every** profile when scanning (merge with `topics/watchlists/<profile_id>.md`).
+
+Use only for handles that are truly cross-positioning. Prefer putting niche-specific radars in the profile file.
+
+| # | platform | handle | niche_tag | why_on_list | weight | status |
+|---|----------|--------|-----------|-------------|--------|--------|
+
+*(empty — add rows only when a handle should appear for all profiles)*
+
+## Id note
+
+Use `#S-01`, `#S-02`, … so shared ids do not collide with profile-local `#1…`.

package/content-engine/starter/topics/watchlists/default.md

@@ -0,0 +1,6 @@
+# Watchlist — profile `default`
+
+Profile-scoped radar. Merge with `topics/watchlists/_shared.md` on every scan.
+
+| # | platform | handle | niche_tag | why_on_list | weight | status |
+|---|----------|--------|-----------|-------------|--------|--------|

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

@@ -0,0 +1,13 @@
+# wiki/
+
+Split library (option B):
+
+| Path | What | Run selection |
+|------|------|----------------|
+| `atoms/` | Techniques per role (hook, cta, proof, …) | Match swipe beats / roles; **none OK** |
+| `claims/` | Falsifiable angles / learned strategies | 0–1 claim; **none OK** |
+
+Selection always starts at each folder’s `_index.md` (Approach B).  
+Filter rows by **platform** and **profiles** (`*` = all profiles) from the current account’s profile_id.  
+Drop atom `dead` and claim `retired`. Prefer atom `working`. Claim `status` is epistemology; rank claims by `win/n`.  
+Atom use history = `published/_index`. Claim proposition evidence = `## Evidence` in the claim file (not in atoms).

package/content-engine/starter/wiki/atoms/_index.md

@@ -0,0 +1,23 @@
+# Atoms index
+
+Techniques per role. Global library; filter with **`profiles`** + **`platforms`** + **`type`**.
+
+**Types:** `hook` | `reframe` | `proof` | `process` | `cta` | `other`  
+**Selection:** table first → drop `dead` → prefer `working` then higher `win/n` → 0–1 per needed type. **`none` allowed.**
+
+## Status (hit-rate, our posts)
+
+`trial` | `working` | `dead` — same meaning as `swipe/_index.md`.  
+`n` / `win` / `loss` = our published cache. Do not share claim’s epistemic words.  
+**No Evidence table in atom files.** Use history = `published/_index` (column `atoms`).
+
+## Ingest
+
+`CLAUDE.md` § Library ingest. Capture offer path A; publish path B + gates.
+
+## Catalog
+
+| id | file | type | platforms | profiles | status | n | win | loss | one_liner |
+|----|------|------|-----------|----------|--------|---|-----|------|-----------|
+
+`profiles: *` = all profiles.

package/content-engine/starter/wiki/atoms/_template.md

@@ -0,0 +1,31 @@
+---
+id: A-NEW
+type: hook                      # hook | reframe | proof | process | cta | other
+status: trial              # trial | working | dead
+platforms: [x]
+profiles: ["*"]
+---
+
+# Title
+
+## Procedure (how to write this beat)
+
+1. …
+
+## Applies when
+
+- …
+
+## Avoid when
+
+- …
+
+## Examples (short; prefer paraphrase over full copy of others)
+
+- …
+
+## Source
+
+- capture / raw / run links if any
+
+<!-- Use history: published/_index column atoms. Do not add ## Evidence here. -->

package/content-engine/starter/wiki/claims/_index.md

@@ -0,0 +1,26 @@
+# Claims index
+
+Falsifiable angles. Global library; filter with **`profiles`** + **`platforms`**.
+
+**Selection:** 0–1 claim per run. Drop `retired`. **`hypothesis` is valid to pick** (that is the test). Rank remaining by `win/n`, not by epistemic status. **`none` allowed.**
+
+## Status (epistemology — “do we still assert this”)
+
+`hypothesis` | `supported` | `weakened` | `retired`
+
+This is **not** hit-rate. Proposition ledger = `## Evidence` in **this claim’s file**.
+
+## Hits (separate)
+
+`n` / `win` / `loss` = our `published.result` when this claim was used. Updating hits **must not** auto-change `status`.
+
+## Ingest / promote
+
+`CLAUDE.md` § Library ingest. No single-win → `supported`. `supported` = ≥3 runs that support the **proposition** + user 批准.
+
+## Catalog
+
+| id | file | type | platforms | profiles | status | n | win | loss | one_liner |
+|----|------|------|-----------|----------|--------|---|-----|------|-----------|
+
+`profiles: *` = all profiles.

package/content-engine/starter/wiki/claims/_template.md

@@ -0,0 +1,38 @@
+---
+id: C-NEW
+version: 1
+type: angle
+status: hypothesis
+platforms: [x]
+profiles: ["*"]
+related_swipe: []
+related_atoms: []
+primary_metric: replies
+confidence: low
+---
+
+# 主张（Objective）
+
+One falsifiable sentence.
+
+## Applies when
+
+- …
+
+## Avoid when
+
+- …
+
+## Procedure (angle constraints for Writer)
+
+1. …
+
+## Evidence
+
+| run | date | result | vs claim | note |
+|-----|------|--------|----------|------|
+|     |      |        |          |      |
+
+## Source
+
+- …

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

@@ -0,0 +1,17 @@
+# workbench/
+
+Read-only observatory. Source of truth remains the markdown indexes.
+
+| File | Role |
+|------|------|
+| `template.html` | UI (do not overwrite on later skill updates if the user customized it) |
+| `build.py` | Scan indexes → `index.html` |
+| `serve.py` | `http://127.0.0.1:8765/workbench/` |
+| `index.html` | Generated — safe to rebuild |
+
+```bash
+python3 workbench/build.py
+python3 workbench/serve.py
+```
+
+If port 8765 is already serving this engine, do not start a second server. Rebuild after Agent writes markdown.