What does this skill do, and when should you use it?
FLOW Blog Workflow is a sub-skill in claude-blog for evidence-led blogging. It exposes the blog-relevant Find, Optimize, and Win stages and uses a prompt library for keyword discovery, intent mapping, gap analysis, content optimization, and conversion evaluation. Its metadata requires Claude Code and Python 3.11+, while prompt synchronization fetches files from the FLOW GitHub repository. It fits Claude Code users who want a structured blog workflow, but it is not documented as a standalone general-purpose writing tool.
On /blog flow, it loads the FLOW framework reference and presents the stage menu; /blog flow find reads five Find prompts and applies them to a topic or URL for demand, intent, and gap signals; /blog flow optimize reviews niche, prior skill output, and URL signals, then selects and loads two or three relevant prompts from 21 Optimize prompts; /blog flow win reads three Win prompts for conversion and BOFU analysis; /blog flow prompts displays the complete index of 30 blog-applicable prompts; /blog flow sync runs python3 scripts/sync_flow.py to fetch prompt files from github.com/AgriciDaniel/flow and display a JSON change summary. Every activation outputs the FLOW CC BY 4.0 attribution notice.
- A blogger has a topic or URL and needs keyword discovery, intent mapping, and content-gap analysis before drafting.
- A SaaS or B2B content team wants a small, context-matched set of optimization prompts instead of loading all 21 prompts.
- A content marketer needs to evaluate BOFU, conversion, and dual-surface performance for a blog URL.
- A skill maintainer needs to synchronize the latest blog prompts from the FLOW repository and inspect added, updated, or unchanged files.
How do you install this skill?
- /blog flow sync fetches and writes new prompt content from GitHub; review diffs and retain a recoverable copy before production use.
- The supplied find prompt files are materially repetitive; the skill description alone does not establish distinct and complete stage coverage.
- Chinese output quality and stable GitHub reachability from mainland China are not demonstrated.
- Treat synced content as untrusted instruction input and review upstream changes before loading it at runtime.
- Shell / CLI
- Network access
- Local filesystem
Claude CodePython 3.11+
This is a sub-skill inside AgriciDaniel/claude-blog, and the source does not document a separate installation. The repository README documents installing the public plugin in Claude Code: /plugin marketplace add AgriciDaniel/claude-blog; /plugin install claude-blog@agricidaniel-claude-blog. After installation, the skill file is skills/blog-flow/SKILL.md.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/AgriciDaniel/claude-blog.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/blog-flow" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
In Claude Code, use /blog flow for the stage menu; /blog flow find <topic-or-url> for discovery analysis; /blog flow optimize <url> for context-matched optimization; /blog flow win <url> for conversion analysis; and /blog flow prompts for the complete prompt index. Use /blog flow sync when reference files need updating; it runs python3 scripts/sync_flow.py and requires Python 3.11+.
What are this skill's strengths and limitations?
- Clearly separates the blog workflow into Find, Optimize, and Win, while keeping the Leverage prompt available through the index.
- The Optimize stage uses niche, previous skill output, and URL signals to select exactly two or three prompts from 21, reducing unnecessary prompt loading.
- The synchronization script supports dry runs, pinned commit SHAs, HTTPS, host allowlisting, response-size limits, and atomic writes.
- The prompt library is attributed under CC BY 4.0, and the skill enforces an attribution notice on every activation.
- It depends on Claude Code /blog commands and skill orchestration, so it is not documented as a drop-in prompt pack for ordinary chat.
- Python 3.11+ is required for synchronization, and synchronization requires network access to GitHub.
- Reference and prompt files are loaded from references/ on demand; missing files require a sync first.
- The supplied source does not document an independent test suite or standalone package for this sub-skill.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| FLOW Blog Workflow this page | 52 · Use with care | ★ 2.4k | 14d ago | MIT |
| GEO Citability Scoring Skill | 48 · Use with care | ★ 11k | 23d ago | MIT |
| Blog Keyword Conflict Detector | 43 · Not recommended | ★ 2.4k | 14d ago | MIT |
| AI Search & GEO Optimizer | 52 · Use with care | ★ 19k | 23d ago | MIT |
| Humanizer — AI Writing Humanizer & Detector Skill | 67 · Recommended | ★ 282 | 1mo ago | MIT |
Compared with improvising blog prompts, FLOW Blog Workflow provides a structured, source-attributed prompt library organized by Find, Optimize, and Win, with context-based selection during optimization. It is less suitable for users who only want a one-off draft quickly.
How did FollowSkills review this skill?
The skill documents HTTPS, host allowlisting, size caps, path-traversal protection, atomic writes, lockfile checks, non-persistence of tokens, and attribution, and limits syncing to an explicit user command. However, syncing downloads content from GitHub and writes files that Claude later loads as instructions; runtime confirmation, rollback, and mandatory isolation against upstream prompt injection are not shown. Eight points are deducted for those gaps. The publisher is unverified and therefore treated as unknown, without an additional penalty.
Command routing, stage behavior, missing-file messages, and sync error handling are described clearly, and optimize-stage context matching is specified. However, several find prompt files are materially duplicated, key referenced files and sync behavior are not fully reproducible from the supplied material, and no blog-flow-specific key-path tests are shown. The score is capped at 8 for static review and reduced for limited coverage and execution evidence.
Triggers, user-invokable status, argument hint, primary scenarios, and the exclusion of local SEO are reasonably clear. Input and output boundaries remain broad, Chinese-language support is not declared, and the core sync function depends on GitHub and other overseas reachability that may be unreliable from mainland China. Seven points are deducted for these limits.
The skill provides frontmatter, MIT licensing, author and version metadata, source attribution, progressive on-demand loading, sync instructions, a lockfile, and troubleshooting guidance. The selected skill does not itself show a changelog, explicit maintenance ownership commitment, or a complete FAQ, and several prompt titles do not distinguish materially different content. Four points are deducted for these governance and documentation gaps.
The skill can provide a FLOW overview, stage selection, prompt indexing, and a staged workflow that requests evidence, measurement, and verification outputs. However, the supplied find prompts are highly repetitive, the actual optimize and win prompt bodies are not included, and no statically verifiable representative end-to-end result is provided. The score is therefore 5.
The material includes source notes, a bibliography with retrieval dates, version metadata, and a sha256 lockfile, and it requires traceable public claims. It does not show blog-flow-specific tests, independent reproduction, or corroborating evidence beyond repository assertions, so the static score is 3.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →