Writing & Content Workflows

A protocol and toolkit for multi-agent collaboration. Agents post turns to a shared board, get each other's messages, challenge each other's work, and record the outcome as a permanent decision. Use when two or more agents (or an agent and a human) need genuine back-and-forth before producing an answer — "have agent X weigh in", "get a second agent to review this", "let the agents confer", "run this by the panel". Ships a zero-infra local backend and a Convex backend for real-time, multi-machine setups.

Stage 1-2 specialization of the Content Engine. Turns a REAL client engagement (a project folder + KG + meeting notes + metrics) into a structured case-study Content Context Brief — problem → before-state → system built → after-state → ONE quantified proprietary result → transferable lesson. This is the proof-of-work artifact at the center of the 2026 LinkedIn authority strategy (see projects/fractal-ai-content/2026-growth-strategy/). Use to turn a construction GC / a crypto OTC desk / a hospitality brand (or any client) into a publishable case study. Hook-agnostic — emits a brief that flows into content-hooks. Triggers — 'case study for', 'turn [client] into a case study', 'build the case study on', 'client teardown', 'content-case-study'.

Stage 3 of the Content Engine. Write title and 3 hook options for an approved content brief. Use when CO hands off an approved brief or you receive a 'Write Title + Hooks' task. Creates Discord thread in content, writes 3 hooks graded A-F against a 4-dimension rubric (psychology / framing / grammar / non-obvious) with reasoning and target emotion, posts for the user to pick or paste his own for a grade-and-rewrite.

Stage 1 of the Content Engine. Diverge then converge on a content seed — generate multiple distinct angles, evaluate them, narrow to one (or a few) winning concept(s). Does NOT write the full brief (that's content-brief). Use when the user has a topic / pain / observation / vague "we should post about X" and the angle is not yet locked.

Stage 6 of the Content Engine. Build lead magnet, create Tally email-capture form, update Notion landing page, and assemble the final post package with CTA link. CO handles infrastructure (GitHub repo, Tally form via API, landing page). CC assembles the final post text.

Stage 4 of the Content Engine. Write the full LinkedIn post from an approved hook. Runs an execute-evaluate loop (max 3 iterations), posts the draft in the Discord thread for the user review, handles up to 2 revision rounds.

Stage 5 of the Content Engine. Create visuals and diagrams for an approved LinkedIn post script. Use when the user approves the script or you receive a 'Create Visuals' task. Creates architecture diagrams, infographics, KG visualizations per brief specs.

Start navigation, get directions and route info via Google Maps app. Tools: intent, http (optional, requires Maps API key for traffic/route details).

Data-driven hook engine. Analyzes the owner's top vs. bottom-performing content to extract winning hook principles, banks them as a persistent grading rubric calibrated to the owner's audience, then generates and grades hooks for new topics. Source-agnostic ingestion (paste, CSV export from LinkedIn/X analytics, or fetch X posts via the scrape skill) with no third-party data platform required. Use when building or refreshing a hook rubric, analyzing why hooks win or lose, generating data-backed hooks for a topic, or grading and rewriting a draft hook. For a quick one-off hook that needs no performance data or grading, use the writing skill instead. Triggers - 'build my hook rubric', 'refresh the hook rubric', 'analyze my best hooks', 'hook machine', 'why do my hooks win', 'generate hooks for', 'grade this hook', 'rewrite this hook'.

Closed-loop social media infographic and carousel generator for Claude Code on the user's Mac. Renders HTML/CSS to PNG via Python Playwright (scripts/render.py), reads the PNG back, evaluates quality, and iterates until polished and publish-ready. Supports LinkedIn carousels, Instagram posts/stories, Twitter/X graphics, and infographics.