The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
refactor: consolidate writing voice rules
Affaan Mustafa635fcbd71566
diff --git a/.agents/skills/article-writing/SKILL.md b/.agents/skills/article-writing/SKILL.mdindex 1a890610a..6cf4339c7 100644--- a/.agents/skills/article-writing/SKILL.md+++ b/.agents/skills/article-writing/SKILL.md@@ -23,50 +23,23 @@ Write long-form content that sounds like an actual person with a point of view,4. Use proof instead of adjectives.5. Never invent facts, credibility, or customer evidence.-## Voice Capture Workflow--If the user wants a specific voice, collect one or more of:-- published articles-- newsletters-- X posts or threads-- docs or memos-- launch notes-- a style guide--Then extract:-- sentence length and rhythm-- whether the writing is compressed, explanatory, sharp, or formal-- how parentheses are used-- how often the writer asks questions-- whether the writer uses fragments, lists, or hard pivots-- formatting habits such as headers, bullets, code blocks, pull quotes-- what the writer clearly avoids+## Voice Handling-If no voice references are given, default to a sharp operator voice: concrete, unsentimental, useful.--## Affaan / ECC Voice Reference+If the user wants a specific voice, run `brand-voice` first and reuse its `VOICE PROFILE`.+Do not duplicate a second style-analysis pass here unless the user explicitly asks for one.-When matching Affaan / ECC voice, bias toward:-- direct claims over scene-setting-- high specificity-- parentheticals used for qualification or over-clarification, not comedy-- capitalization chosen situationally, not as a gimmick-- very low tolerance for fake thought-leadership cadence-- almost no bait questions+If no voice references are given, default to a sharp operator voice: concrete, unsentimental, useful.## Banned PatternsDelete and rewrite any of these:- "In today's rapidly evolving landscape"- "game-changer", "cutting-edge", "revolutionary"-- "no fluff"-- "not X, just Y"- "here's why this matters" as a standalone bridge- fake vulnerability arcs- a closing question added only to juice engagement-- forced lowercase-- corny parenthetical asides- biography padding that does not move the argument+- generic AI throat-clearing that delays the point## Writing Process@@ -101,6 +74,6 @@ Delete and rewrite any of these:Before delivering:- factual claims are backed by provided sources- generic AI transitions are gone-- the voice matches the supplied examples+- the voice matches the supplied examples or the agreed `VOICE PROFILE`- every section adds something new- formatting matches the intended mediumdiff --git a/.agents/skills/content-engine/SKILL.md b/.agents/skills/content-engine/SKILL.mdindex d1678a1df..4467724ae 100644--- a/.agents/skills/content-engine/SKILL.md+++ b/.agents/skills/content-engine/SKILL.md@@ -38,50 +38,28 @@ Before drafting, identify the source set:If the user wants a specific voice, build a voice profile from real examples before writing.Use `brand-voice` as the canonical workflow when voice consistency matters across more than one output.-## Voice Capture Workflow+## Voice Handling-Run `brand-voice` first when:+`brand-voice` is the canonical voice layer.++Run it first when:- there are multiple downstream outputs- the user explicitly cares about writing style- the content is launch, outreach, or reputation-sensitive-At minimum, produce a compact `VOICE PROFILE` covering:-- rhythm-- compression-- capitalization-- parenthetical use-- question use-- preferred moves-- banned moves--Do not start drafting until the voice profile is clear enough to enforce.--## Affaan / ECC Voice Reference--When the user wants Affaan / ECC voice specifically, default to this unless newer examples clearly override it:-- direct, compressed, concrete-- strong preference for specific claims, numbers, mechanisms, and receipts-- parentheticals used to qualify, narrow, or over-clarify, not to do corny bits-- lowercase is optional, not mandatory-- questions are rare and should not be added as bait-- transitions should feel earned, not polished-- tone can be sharp or blunt, but should not sound like a content marketer+Reuse the resulting `VOICE PROFILE` here instead of rebuilding a second voice model.+If the user wants Affaan / ECC voice specifically, still treat `brand-voice` as the source of truth and feed it the best live or source-derived material available.## Hard BansDelete and rewrite any of these:- "In today's rapidly evolving landscape"- "game-changer", "revolutionary", "cutting-edge"-- "no fluff"-- "not X, just Y"- "here's why this matters" unless it is followed immediately by something concrete-- "Excited to share"-- fake curiosity gaps- ending with a LinkedIn-style question just to farm replies-- forced lowercase when the source voice does not call for it- forced casualness on LinkedIn-- parenthetical jokes that were not present in the source voice+- fake engagement padding that was not present in the source material## Platform Adaptation Rulesdiff --git a/.agents/skills/crosspost/SKILL.md b/.agents/skills/crosspost/SKILL.mdindex 2c595fb96..e94df3c31 100644--- a/.agents/skills/crosspost/SKILL.md+++ b/.agents/skills/crosspost/SKILL.md@@ -39,13 +39,8 @@ Use `content-engine` first if the source still needs voice shaping.Run `brand-voice` first if the source voice is not already captured in the current session.-Before adapting, note:-- how blunt or explanatory the source is-- whether the source uses fragments, lists, or longer transitions-- whether the source uses parentheses-- whether the source avoids questions, hashtags, or CTA language--The adaptation should preserve that fingerprint.+Reuse the resulting `VOICE PROFILE` directly.+Do not build a second ad hoc voice checklist here unless the user explicitly wants a fresh override for this campaign.### Step 3: Adapt by Platform Constraint@@ -91,7 +86,6 @@ Delete and rewrite any of these:- "Here's what I learned"- "What do you think?"- "link in bio" unless that is literally true-- "not X, just Y"- generic "professional takeaway" paragraphs that were not in the source## Output Formatdiff --git a/.agents/skills/investor-outreach/SKILL.md b/.agents/skills/investor-outreach/SKILL.mdindex 9abef5214..f34860e42 100644--- a/.agents/skills/investor-outreach/SKILL.md+++ b/.agents/skills/investor-outreach/SKILL.md@@ -24,6 +24,11 @@ Write investor communication that is short, concrete, and easy to act on.4. Stay concise.5. Never send copy that could go to any investor.+## Voice Handling++If the user's voice matters, run `brand-voice` first and reuse its `VOICE PROFILE`.+This skill should keep the investor-specific structure and ask discipline, not recreate its own parallel voice system.+## Hard BansDelete and rewrite any of these:@@ -31,7 +36,6 @@ Delete and rewrite any of these:- "excited to share"- generic thesis praise without a real tie-in- vague founder adjectives-- "no fluff"- begging language- soft closing questions when a direct ask is clearerdiff --git a/WORKING-CONTEXT.md b/WORKING-CONTEXT.mdindex 7fb24edcf..2cc027cda 100644--- a/WORKING-CONTEXT.md+++ b/WORKING-CONTEXT.md@@ -111,3 +111,4 @@ Keep this file detailed for only the current sprint, blockers, and next actions.- 2026-04-01: Added `connections-optimizer` as the review-first social-graph reorganization workflow for X and LinkedIn, with explicit pruning modes, browser fallback expectations, and Apple Mail drafting guidance.- 2026-04-01: Added `manim-video` as the reusable technical explainer lane and seeded it with a starter network-graph scene so launch and systems animations do not depend on one-off scratch scripts.- 2026-04-02: Re-extracted `social-graph-ranker` as a standalone primitive because the weighted bridge-decay model is reusable outside the full lead workflow. `lead-intelligence` now points to it for canonical graph ranking instead of carrying the full algorithm explanation inline, while `connections-optimizer` stays the broader operator layer for pruning, adds, and outbound review packs.+- 2026-04-02: Applied the same consolidation rule to the writing lane. `brand-voice` remains the canonical voice system, while `content-engine`, `crosspost`, `article-writing`, and `investor-outreach` now keep only workflow-specific guidance instead of duplicating a second Affaan/ECC voice model or repeating the full ban list in multiple places.diff --git a/skills/article-writing/SKILL.md b/skills/article-writing/SKILL.mdindex 1a890610a..6cf4339c7 100644--- a/skills/article-writing/SKILL.md+++ b/skills/article-writing/SKILL.md@@ -23,50 +23,23 @@ Write long-form content that sounds like an actual person with a point of view,4. Use proof instead of adjectives.5. Never invent facts, credibility, or customer evidence.-## Voice Capture Workflow--If the user wants a specific voice, collect one or more of:-- published articles-- newsletters-- X posts or threads-- docs or memos-- launch notes-- a style guide--Then extract:-- sentence length and rhythm-- whether the writing is compressed, explanatory, sharp, or formal-- how parentheses are used-- how often the writer asks questions-- whether the writer uses fragments, lists, or hard pivots-- formatting habits such as headers, bullets, code blocks, pull quotes-- what the writer clearly avoids+## Voice Handling-If no voice references are given, default to a sharp operator voice: concrete, unsentimental, useful.--## Affaan / ECC Voice Reference+If the user wants a specific voice, run `brand-voice` first and reuse its `VOICE PROFILE`.+Do not duplicate a second style-analysis pass here unless the user explicitly asks for one.-When matching Affaan / ECC voice, bias toward:-- direct claims over scene-setting-- high specificity-- parentheticals used for qualification or over-clarification, not comedy-- capitalization chosen situationally, not as a gimmick-- very low tolerance for fake thought-leadership cadence-- almost no bait questions+If no voice references are given, default to a sharp operator voice: concrete, unsentimental, useful.## Banned PatternsDelete and rewrite any of these:- "In today's rapidly evolving landscape"- "game-changer", "cutting-edge", "revolutionary"-- "no fluff"-- "not X, just Y"- "here's why this matters" as a standalone bridge- fake vulnerability arcs- a closing question added only to juice engagement-- forced lowercase-- corny parenthetical asides- biography padding that does not move the argument+- generic AI throat-clearing that delays the point## Writing Process@@ -101,6 +74,6 @@ Delete and rewrite any of these:Before delivering:- factual claims are backed by provided sources- generic AI transitions are gone-- the voice matches the supplied examples+- the voice matches the supplied examples or the agreed `VOICE PROFILE`- every section adds something new- formatting matches the intended mediumdiff --git a/skills/content-engine/SKILL.md b/skills/content-engine/SKILL.mdindex d1678a1df..4467724ae 100644--- a/skills/content-engine/SKILL.md+++ b/skills/content-engine/SKILL.md@@ -38,50 +38,28 @@ Before drafting, identify the source set:If the user wants a specific voice, build a voice profile from real examples before writing.Use `brand-voice` as the canonical workflow when voice consistency matters across more than one output.-## Voice Capture Workflow+## Voice Handling-Run `brand-voice` first when:+`brand-voice` is the canonical voice layer.++Run it first when:- there are multiple downstream outputs- the user explicitly cares about writing style- the content is launch, outreach, or reputation-sensitive-At minimum, produce a compact `VOICE PROFILE` covering:-- rhythm-- compression-- capitalization-- parenthetical use-- question use-- preferred moves-- banned moves--Do not start drafting until the voice profile is clear enough to enforce.--## Affaan / ECC Voice Reference--When the user wants Affaan / ECC voice specifically, default to this unless newer examples clearly override it:-- direct, compressed, concrete-- strong preference for specific claims, numbers, mechanisms, and receipts-- parentheticals used to qualify, narrow, or over-clarify, not to do corny bits-- lowercase is optional, not mandatory-- questions are rare and should not be added as bait-- transitions should feel earned, not polished-- tone can be sharp or blunt, but should not sound like a content marketer+Reuse the resulting `VOICE PROFILE` here instead of rebuilding a second voice model.+If the user wants Affaan / ECC voice specifically, still treat `brand-voice` as the source of truth and feed it the best live or source-derived material available.## Hard BansDelete and rewrite any of these:- "In today's rapidly evolving landscape"- "game-changer", "revolutionary", "cutting-edge"-- "no fluff"-- "not X, just Y"- "here's why this matters" unless it is followed immediately by something concrete-- "Excited to share"-- fake curiosity gaps- ending with a LinkedIn-style question just to farm replies-- forced lowercase when the source voice does not call for it- forced casualness on LinkedIn-- parenthetical jokes that were not present in the source voice+- fake engagement padding that was not present in the source material## Platform Adaptation Rulesdiff --git a/skills/crosspost/SKILL.md b/skills/crosspost/SKILL.mdindex 2c595fb96..e94df3c31 100644--- a/skills/crosspost/SKILL.md+++ b/skills/crosspost/SKILL.md@@ -39,13 +39,8 @@ Use `content-engine` first if the source still needs voice shaping.Run `brand-voice` first if the source voice is not already captured in the current session.-Before adapting, note:-- how blunt or explanatory the source is-- whether the source uses fragments, lists, or longer transitions-- whether the source uses parentheses-- whether the source avoids questions, hashtags, or CTA language--The adaptation should preserve that fingerprint.+Reuse the resulting `VOICE PROFILE` directly.+Do not build a second ad hoc voice checklist here unless the user explicitly wants a fresh override for this campaign.### Step 3: Adapt by Platform Constraint@@ -91,7 +86,6 @@ Delete and rewrite any of these:- "Here's what I learned"- "What do you think?"- "link in bio" unless that is literally true-- "not X, just Y"- generic "professional takeaway" paragraphs that were not in the source## Output Formatdiff --git a/skills/investor-outreach/SKILL.md b/skills/investor-outreach/SKILL.mdindex 9abef5214..f34860e42 100644--- a/skills/investor-outreach/SKILL.md+++ b/skills/investor-outreach/SKILL.md@@ -24,6 +24,11 @@ Write investor communication that is short, concrete, and easy to act on.4. Stay concise.5. Never send copy that could go to any investor.+## Voice Handling++If the user's voice matters, run `brand-voice` first and reuse its `VOICE PROFILE`.+This skill should keep the investor-specific structure and ask discipline, not recreate its own parallel voice system.+## Hard BansDelete and rewrite any of these:@@ -31,7 +36,6 @@ Delete and rewrite any of these:- "excited to share"- generic thesis praise without a real tie-in- vague founder adjectives-- "no fluff"- begging language- soft closing questions when a direct ask is clearer