HOME SKILLS BLOG GITHUB
// SKILL

FLOW FRAMEWORK
EVIDENCE-LED BLOGGING

Stop running improvised AI queries against your blog workflow. FLOW gives Claude Blog 30 evidence-led prompts organized across three stages: Find, Optimize, Win. Every analysis is anchored in a structured prompt with stated evidence, not vibes. Bundled under CC BY 4.0. Free with Claude Blog.

FRAMEWORK © DANIEL AGRICI · CC BY 4.0 · github.com/AgriciDaniel/flow
$
/blog flow [stage]

REQUIRES CLAUDE BLOG INSTALLED IN CLAUDE CODE

// HOW IT WORKS

3 STAGES. 30 PROMPTS.

FLOW is an operating model for the AI-search era. Claude Blog exposes the three blog-relevant stages (Find, Optimize, Win) and keeps the single Leverage prompt available through the prompts index. Each stage ships with structured prompts that Claude Blog surfaces on demand, scoped to the situation you describe.

Instead of guessing how to query an LLM about your content, you pick the stage you are in and FLOW selects the right prompts. The Optimize stage uses context-matching orchestration: with 21 prompts available, Claude Blog loads only the 2-3 most relevant based on your niche, URL signals, and prior skill output already in your conversation (like a recent /blog analyze score).

F
FIND
Keyword discovery, intent mapping, gap analysis. 5 prompts. Use when starting a new topic and you do not yet know what to write. Hands off to /blog brief and /blog outline downstream.
O
OPTIMIZE
21 prompts covering on-page, technical, content quality, freshness, structured data. Claude Blog selects the 2-3 most relevant per analysis, never dumps all 21. Hands off to /blog rewrite, /blog seo-check, /blog geo, /blog schema, /blog factcheck.
W
WIN
Bottom-of-funnel and conversion. 3 prompts. Use on money pages, product reviews, and pillar pages to convert organic traffic into pipeline. Hands off to /blog repurpose and /blog audio.
Σ
CONTEXT-MATCHED
The Optimize stage never dumps all 21 prompts. It selects 2-3 based on niche, URL signals, and prior skill output (analyze score deltas, factcheck flags, geo readiness). Surface area without noise.
// USAGE

HOW TO RUN FLOW

Stage Commands

CommandWhat it does
/blog flowShows the stage menu with one-line descriptions
/blog flow find [topic|url]Runs all 5 Find prompts for keyword + intent discovery
/blog flow optimize [url]Selects and runs 2-3 of 21 Optimize prompts by context
/blog flow win [url]Runs all 3 Win prompts for BOFU and conversion
/blog flow promptsLists every prompt with stage, name, and trigger
/blog flow syncPulls the latest prompts from github.com/AgriciDaniel/flow

Cross-Skill References

FLOW does not duplicate the rest of Claude Blog. It points at the right specialist skill when one is needed:

Why the Local Stage is Excluded

The full FLOW framework includes 11 Local-stage prompts targeting Google Business Profile, NAP consistency, citations, and map-pack visibility. Those prompts target brick-and-mortar work, not blog content. For local SEO use the /seo local skill from the claude-seo plugin instead.

Syncing the Prompt Library

FLOW prompts live in the upstream repo at github.com/AgriciDaniel/flow. Run /blog flow sync to pull the latest version. The sync script is anonymous-first against the GitHub API (no token required for under 60 requests per hour on public repos). A SHA-256 lockfile (flow-prompts.lock) tracks baseline integrity and reports upstream drift before writing.

Attribution

Every /blog flow activation outputs this header before analysis:

Framework and prompts © Daniel Agrici, CC BY 4.0, github.com/AgriciDaniel/flow

The attribution is required by the CC BY 4.0 license and is automated by the sync script. Do not remove it.

// FAQ

QUESTIONS ABOUT FLOW

FLOW is an evidence-led operating model built for the AI-search era. For blogs it organizes work into three stages: Find (keyword discovery, intent mapping, gap analysis), Optimize (on-page, technical, content quality, freshness, structured data), and Win (BOFU, conversion, dual-surface scorecard). Each stage ships with structured AI prompts that drive analysis based on evidence rather than improvised queries.
30 blog-applicable prompts total: 5 Find prompts, 21 Optimize prompts, and 3 Win prompts, plus 1 Leverage prompt accessible via the full prompts index. The Optimize stage uses context-matching orchestration: instead of dumping all 21 prompts, Claude Blog selects 2-3 most relevant based on niche, URL signals, and any prior skill output (like a recent /blog analyze score) in the conversation.
Yes. The FLOW prompts are released under CC BY 4.0 by Daniel Agrici and bundled inside Claude Blog with attribution. Claude Blog's skill code remains MIT-licensed. You can use, modify, and redistribute the prompts as long as attribution to the FLOW repo (github.com/AgriciDaniel/flow) is preserved.
FLOW points to the right specialist skill at every stage rather than duplicating their logic. Find stage hands off to /blog brief and /blog outline. Optimize stage hands off to /blog rewrite, /blog seo-check, /blog geo, /blog schema, and /blog factcheck. Win stage hands off to /blog repurpose and /blog audio. The framework sits on top of the existing 27 skills rather than replacing any of them.
The full FLOW framework includes 11 Local-stage prompts targeting Google Business Profile, NAP consistency, citations, and map-pack visibility. Those prompts are intentionally excluded from blog-flow because they target brick-and-mortar local SEO, not blog content workflows. For local SEO use /seo local from the claude-seo plugin instead.
// RELATED SKILLS

EXPLORE MORE

VIEW ALL 28 SKILLS →

RUN YOUR FIRST FLOW LOOP
IN 30 SECONDS.

$
git clone --depth 1 https://github.com/AgriciDaniel/claude-blog.git && bash claude-blog/install.sh
VIEW ON GITHUB ALL SKILLS >