@sellable/mcp 0.1.237 → 0.1.239

package/skills/create-post/references/post-validation.md

@@ -12,9 +12,12 @@ Every saved draft needs a validation receipt. A draft without this receipt is no
 - `proofClaimsUsed`
 - `proofClaimSources`
 - `storyFilesConsulted`
+- `goldStandardsConsulted`
 - `voiceRulesApplied`
+- `linkedinPreviewAudit`
 - `simplifierConcreteLanguageAudit`
 - `antiAiAudit`
+- `outboundAiTellAudit`
 - `finalizerChanges`
 - `blockedReasons`
 - `retryNeededReasons`
@@ -29,6 +32,10 @@ Each candidate should include:
 - hook text
 - source pattern
 - score
+- char count including newlines and first-line / first-two-line preview measurements
+- physical line count, content line count, longest nonblank line, and blank-line risk
+- `previewBudgetStatus`: `pass`, `warn`, or `fail`
+- mobile and desktop preview fit
 - proof/story dependency
 - AI-tell risk
 - why it should win or lose
@@ -43,6 +50,48 @@ After the first draft:
 4. replace generic language with concrete words only when supported
 5. preserve the user's actual story and point
 6. remove AI tells
+7. re-check LinkedIn preview fit after edits
+
+## LinkedIn Preview Audit
+
+Audit the selected hook and top candidates against conservative LinkedIn
+preview budgets. LinkedIn does not publish exact "see more" cutoff rules, and
+rendering varies by device, app version, font, media, and line break. This audit
+is a mobile-first safety gate, not a claim about an official LinkedIn limit.
+
+Use:
+
+- `pass`: hook is <= 110 chars including newlines, every nonblank line is <= 45 chars, and the hook's core point lands before likely truncation.
+- `warn`: hook is 111-140 chars including newlines, any nonblank line is 46-55 chars, or blank lines create visual-line risk. Blank lines are allowed, but they count as physical lines.
+- `fail`: hook is > 140 chars including newlines, any nonblank line is > 55 chars, or the hook's point depends on text after likely truncation.
+
+Desktop preview usually has more room. Still record desktop fit, but never let
+desktop fit compensate for a mobile `fail`.
+
+Record:
+
+- `charCount`
+- `charCountIncludingNewlines`
+- `firstLineChars`
+- `firstTwoLinesChars`
+- `physicalLineCount`
+- `contentLineCount`
+- `longestNonblankLineChars`
+- `blankLineCountBeforeFold`
+- `blankLineVisualRisk`
+- `corePointBeforeLikelyTruncation`
+- `previewBudgetStatus`
+- `mobilePreviewFit`
+- `desktopPreviewFit`
+- `lineCountEstimate`
+- `truncationRisk`
+- `rewriteIfTruncated`
+- `compactFallback` when `previewBudgetStatus` is `warn`
+
+If the hook only works after likely truncation, rewrite it. A draft cannot be
+`ready` with `previewBudgetStatus: fail`. A draft may be `ready` with
+`previewBudgetStatus: warn` only when the warning is explicit, usually because
+the user prefers blank-line rhythm, and the receipt includes a compact fallback.
 
 ## Simplifier / Concrete-Language Audit
 
@@ -59,7 +108,25 @@ Do not make language concrete by inventing new facts.
 
 ## Anti-AI Audit
 
-Reject or rewrite:
+Start with the Outbound Base AI-Tell Layer from the message pipeline, adapted for public posts. This is a starter smell detector, not permission to run outbound workflows or force outbound copy style.
+
+Reject or rewrite outbound-derived tells when they appear in a post:
+
+- body-level self-introductions that add no story context
+- source-citation phrasing such as "your bio says", "your profile shows", or "saw on LinkedIn"
+- explicit date or duration precision that feels scraped
+- block quotes of another person's phrasing longer than 4 words
+- parenthetical stack or feature dumps
+- over-precise numerics without a traceable source
+- hardcoded signoffs
+- vague proof brags such as "trusted by leaders" without named or sourced proof
+- formal CTA phrasing
+- reusable openers that could fit any topic
+- em dashes in the draft body unless the user's post rules explicitly allow them
+- resume-recap or scraped-personalization bridges
+- flattened "worth sending" bridges
+
+Also reject or rewrite post-native AI tells:
 
 - "game-changing"
 - "leverage"
@@ -67,9 +134,13 @@ Reject or rewrite:
 - "this is key"
 - "let that sink in"
 - "read that again"
+- "here's the thing"
+- "unpopular opinion" without a specific argument
 - generic lessons without a story
 - tidy three-part frameworks that were not in the source material
 - fabricated numbers or examples
+- contrarian bait without proof
+- lead-magnet bait when the user did not ask for a lead magnet
 
 ## Proof And Voice Gates
 
@@ -83,6 +154,10 @@ Every claim must trace to at least one of:
 - `core/answer-bank.md`
 - approved reference material
 
+Gold standards can guide structure, rhythm, and judgment. They cannot create new
+claims. If an outside gold standard is used, verify the draft copied no outside
+wording unless the user explicitly approved quotation.
+
 If the necessary proof or story is missing, ask the user or return blocked/retry-needed.
 
 ## Ready Status

package/skills/generate-messages/SKILL.md

@@ -45,6 +45,11 @@ message assets through Sellable MCP tools:
 After candidate generation and revision, and before returning `ready`, load
 `create-campaign-v2-validation` through Sellable MCP as the final gate.
 
+Both live and dry runs must produce or validate `message-validation.md` with
+the same quality gates. The selected winner must pass the `Concrete Language Audit`:
+identify and replace or cut abstract verbs, abstract nouns, abstract adjectives,
+abstract adverbs, and cliches before the message is treated as ready.
+
 Never reconstruct that branch from `brief.md`, `lead-review.md`,
 `lead-sample.json`, `lead-filter.md`, `message-validation.md`, local files, or
 direct database reads. If any required message asset cannot be loaded through MCP

package/skills/interview/references/reference-curation.md

@@ -66,3 +66,30 @@ Personal LinkedIn posts and inspiration examples deserve especially clear
 notes: explain whether the lesson is voice, hook shape, proof handling,
 argument structure, audience fit, or a pattern to avoid. Do not mix admiration
 with permission to copy; many inspiration examples should be structural only.
+
+## LinkedIn Post Gold Standards
+
+The approved LinkedIn post gold-standard pack lives in
+`~/.sellable/configs/core/references/linkedin-posts/` and is capped at 20
+approved examples. User-owned examples can be personal best posts or
+user-approved space benchmarks researched from high-performing posts in the
+space.
+
+Do not add researched outside posts directly to the approved pack. First show a
+candidate list and ask which examples the user wants to add, skip, mark as
+anti-examples, or use to replace an existing item when the pack is full.
+
+Each approved post reference should split the lesson into:
+
+- hook mechanism
+- content/body mechanism
+- rhythm and pacing
+- sentence structure
+- proof/story use
+- voice fit
+- risks and allowed use
+
+Outside examples are structure/taste references only unless the user explicitly
+approves quotation. Personal posts can be stronger voice references, but still
+record whether to quote, imitate structure, summarize, keep private, or treat as
+an anti-example.