| name | blog-post-writer |
| description | Orchestrate blog writing from brief to final draft using the repo blog corpus, lightweight web research, One Horizon context docs, image planning, two writing passes, and the required review sequence. Use when Codex needs to draft or revise an original long-form blog article for this repo. Do not use this skill for LinkedIn posts, social-first writing, or storage-only tasks. |
Blog Post Writer
Overview
Ground every article in the selected Ink profile's published blog corpus resolved from that profile's blogPublishingConfig. First confirm what general type of blog post the user wants, then find related internal articles for angle, structure, and voice, not for reusable sourcing. Build a fresh external source set, validate an outline, and write long-form copy in two passes that fits the thesis, the confirmed article type, the audience, and the desired next step.
Use the workflow in references/workflow.md as the authoritative sequence for this skill. Every workflow step is mandatory. Do not skip, collapse, reorder, or silently defer any step. If a step has no substantive output, record that outcome explicitly instead of omitting the step.
Before presenting anything as final, complete and report a visible review ledger. Silent completion does not count. If the Ramsay line is missing, or its Must Fix items are still open, the article is still a working draft.
Quick Start
- Read
references/workflow.md.
- If the brief includes partner, referral, commerce, or affiliate links, read
references/affiliate-integration.md before outlining.
- Load the minimum author-scoped One Horizon context docs (resolved in step 10 below; read here only if the blog post type requires personal context before the archive search).
- Resolve the active Ink profile, then read the selected profile's
blogPublishingConfig.
- Ask the user to confirm the general blog post type before drafting. If the brief already implies one, state the inferred type and ask the user to confirm or correct it. If the user says something broad like
article, map it to the nearest playbook in references/format-playbooks.md and verify the mapped type back to the user. Do not start internal research, external research, or outline work until the type is confirmed.
Exception: when this skill is invoked by ../content-creation-runner/SKILL.md or ../content-publishing-runner/SKILL.md and the One Horizon brief, stored draft, comments, or handoff already provide a confirmed or normalized type, treat that as the confirmed type and continue without asking the user again.
- If the profile blog publishing file is missing, or if either stored folder is
[unset] or missing on disk, ask the user for the existing blog articles folder and the published blog output folder, then update the selected profile's blogPublishingConfig before continuing.
- Search
source_articles_dir from the selected profile's blog publishing config for up to 3-7 relevant internal examples. Prefer the same theme, argument shape, or confirmed article type first. Use them for voice and structure only. Do not inherit their source lists or citations. If none are relevant, say so explicitly and continue.
- Read
../linkedin-social-writer/references/mcp-tools.md.
- Do a lightweight external research pass on the topic. Build a fresh source set for the article with the local MCP research tools.
- Resolve the author in the selected profile workspace and load only the relevant author-scoped One Horizon context docs. Use
../one-horizon-context-setup/references/context-doc-templates.md for the naming and missing-doc contract.
- Distill voice markers with
references/style-capture.md.
- Draft an outline and image plan and validate both against
references/format-playbooks.md and the brief constraints.
- Use
search_images then download_image from ../blog-image-finder/SKILL.md for final external blog assets.
- Write pass one by locking the opening, section spine,
--- section breaks, cover image, and inline image placements.
- Write pass two by adding proof, examples, transitions, detail, and the final inline images while keeping the article prose-led rather than list-led.
- Run every mandatory review pass described in
references/review-passes.md, including ../blog-post-ramsay-review/SKILL.md after tone review, then report the review ledger with Ramsay verdict, score, and Must Fix disposition.
- Save unpublished drafts in the selected profile's
contentRoots.blogDrafts if the user wants them persisted.
Working Agreement
- Resolve the active Ink profile before One Horizon lookups, corpus search, blog path reads, or draft storage. If multiple profiles exist and none is named, ask which profile to use.
- Use the selected profile's blog corpus as the source of truth for tone, pacing, argument style, section shape, CTA intensity, and how directly to sell.
- Read the selected profile's
blogPublishingConfig for the active path values.
- Use One Horizon context docs from the selected profile workspace for live runtime context. Do not use tracked repo files for live context. The exception is the selected profile's
blogPublishingConfig for machine-local blog path state.
- If a required author-scoped context doc is missing or unusable, use
../one-horizon-context-setup/SKILL.md to create the missing doc through its confirmation flow before drafting.
- If a One Horizon tool call is missing or fails, follow
../one-horizon-context-setup/references/mcp-readiness.md. Do not search tracked repo files as a substitute for live context.
- Always ask the user to confirm the general blog post type before outlining. If the type is implied, reflect the inferred type and ask for confirmation or correction instead of silently assuming.
- When the input comes from
content-creation-runner or content-publishing-runner and the One Horizon handoff already includes a confirmed or normalized type, treat that handoff type as sufficient confirmation. Do not ask the user to reconfirm it unless the stored type is missing or internally inconsistent.
- Normalize broad labels such as
article, story, or essay to the closest playbook in references/format-playbooks.md, then verify the mapped type with the user before continuing.
- Use the exact confirmation and verification wording from
references/format-playbooks.md when the type is missing or only implied.
- If the local file is missing, create it after asking the user for the required folders.
- If a stored folder no longer exists, ask the user for the new folder and update the selected profile's
blogPublishingConfig before continuing.
- Do not assume
content/blog/posts/.
- Research internally first so the archive shapes the angle before outside sources do, but do not treat earlier articles as a reusable source bank.
- Do not reuse or carry forward source lists, citations, or outbound links from other articles. If an older piece points to something useful, treat it as a lead and re-open the original source directly before deciding whether it belongs in the new draft.
- Follow with a brief web research pass to sharpen the angle, gather current terminology, and surface claims that need verification later.
- Use One Horizon context docs from the selected profile workspace to make the writing authentic, but load only what is relevant to the article.
- Reuse patterns, not sentences. Do not remix published copy line by line.
- Prefer a narrow, evidence-based voice profile over generic SEO-blog language.
- For blog articles, aim for a sharper, more reportorial tech-magazine voice, closer to Verge-style analysis than to a generic SaaS post. Borrow the energy, clarity, and sense of stakes, not any outlet's phrasing.
- Prefer primary sources, company docs, and strong secondary reporting over generic SEO summaries when researching.
- Always draft an outline before prose and validate it against the active format playbook and the brief.
- Use the confirmed article type to pick internal examples, the relevant One Horizon context docs, and the matching format playbook.
- Every workflow step is mandatory. Execute the step even when the result is
no relevant internal examples, no external sources materially changed the draft, or no URLs were present to validate.
- Make skipped-looking work explicit. If a mandatory step or pass ran but made no material change, record that outcome explicitly instead of omitting the step from the response.
- Always produce one cover image for
metadata.coverImage.
- Always plan at least one inline image beyond the cover image.
- Follow the section-coverage rule in
references/workflow.md for inline image count and placement.
- Use
search_images then download_image from blog-image-finder for final external blog assets, passing the selected Ink profile when it is known. Do not use search_unsplash, generic web search, or raw image URLs as the final asset path.
- Keep
metadata.coverImage as a plain posts/... asset path such as posts/my-image-name.jpg.
- For inline blog images, use
getImageUrl('posts/my-image-name.jpg') as the final URL in the article.
- Keep article image paths separate from storage upload targets. The article contract stays
posts/..., while uploaded objects should land under images/posts/....
- Treat
getImageUrl() as the fixed project contract for blog image URLs. Do not inspect, explain, or reverse-engineer how it works unless the user explicitly asks.
- Form
search_images.query as a short visual concept, not a literal description of the article thesis. Search for simple scenes like empty office or quiet workspace, then narrow only if needed.
- Assume the image MCP tools work. Do not run verifier or setup checks before using them. Only troubleshoot MCP after a concrete tool failure.
- Never include placeholder asset manifests, JSX comments, HTML comments, TODO notes, or editorial notes inside the article file. Do not add blocks like
{/* Placeholder assets expected ... */}.
- If asset work is still pending, keep that note in the response or handoff summary, not in the
.mdx body.
- Use
upload_image from blog-image-uploader only when the user wants the selected assets uploaded to the selected profile's blog image bucket or the publish workflow requires it. Pass the selected Ink profile. When uploading, target images/posts/..., not posts/....
- Treat pass one as structure work and pass two as expansion work. Do not try to fully solve the whole article in one run.
- Use the exact workflow order. Do not skip or merge mandatory steps, and do not start full prose before the outline is validated.
- If the article includes affiliate links, plan them during the outline and source work using
references/affiliate-integration.md; do not append them after the article is otherwise finished.
- Omit affiliate links when there is no natural buying-decision moment in the article.
- Put affiliate disclosures close to the links and in plain language. Do not bury commercial relationships in a footer, source list, or detached note.
- Keep affiliate claims modest and source-backed. Avoid unsupported
best, savings, medical, performance, or guaranteed-outcome claims.
- Do not skip internal research when relevant published blog examples exist.
- Default to paragraphs. Avoid bullet points and numbered lists in the article body unless the brief or the material truly leaves no good prose alternative.
- Separate major sections with
---.
- Use
Final draft, ready to publish, or equivalent only when the review ledger is complete and the Ramsay Must Fix status is none, addressed, or waived by user. Otherwise label the result as a working draft.
- Treat visuals, code samples, screenshots, charts, and citations as part of the brief. Make the text and the asset work together.
- Ask only the smallest set of high-leverage questions. The blog-type confirmation is mandatory; beyond that, if the brief, corpus, and One Horizon context are sufficient, do not ask.
- Keep unpublished drafts separate from the selected profile's published corpus at all times.
- Published blog posts must have a unique
metadata.date within the selected profile's publish_output_dir. Before saving or returning publish-ready MDX, scan existing published posts, pick an unused date, and keep the filename date and metadata.date identical. Do not solve published-date collisions with -2, -3, or same-day duplicate dates.
Workflow
Follow the shared blog workflow in references/workflow.md.
For articles:
- Confirm the blog post type with the user before internal research and outline work.
- Use the matching article-type playbook in
references/format-playbooks.md, normalizing broad labels to the nearest playbook and verifying the mapping when needed.
- Do not treat a casual handoff suggestion or inferred label as confirmed. Approved One Horizon Blog ideas are the exception: their stored
Blog type is confirmed once the idea is moved to Planned, unless the reviewer changed it.
- Research the published archive before the web.
- Build a fresh source set for each article instead of carrying over citations from older posts.
- Draft and validate an outline plus image plan before writing full prose.
- Require one
coverImage in metadata and at least one inline image in the article body.
- Follow the inline-image coverage rule in
references/workflow.md for longer articles.
- Use at least 2 writing passes before the review loop.
- Keep blog articles prose-led, not list-led, unless the brief explicitly demands a list.
- Use
--- between major sections in the article body.
- Start with
../content-humanizer/SKILL.md.
- Continue with
../content-style-review/SKILL.md.
- Continue with
../fact-check/SKILL.md.
- Continue with
../source-url-check/SKILL.md for every article. If there are no URLs to validate, say so explicitly instead of skipping the pass.
- If the draft includes affiliate links and
affiliate-compliance-review is available in the session, run it before tone review and resolve required edits before Ramsay review.
- Continue with
../content-tone-review/SKILL.md as the last content-adjustment pass before Ramsay review.
- Continue with
../blog-post-ramsay-review/SKILL.md for every article after tone review. If that pass returns any Must Fix items, address all of them before calling the draft final unless the user explicitly waives one, then verify the fixes against the Ramsay findings instead of assuming they are resolved.
- Finish with a
Review Ledger section that lists all standard six passes in order, plus Affiliate compliance review before tone review when affiliate links are present. For Ramsay review, include the verdict, the score, and whether Must Fix is none, addressed, or waived by user.
- Add one extra self-edit for longer articles, argument-heavy articles, or drafts with multiple claims.
- Do not present a draft as final until every mandatory step and pass above has been completed, reported in the review ledger, and accounted for.
Format Rules
article: Deliver one polished article in the repo's blog voice. Offer multiple title options only if asked or if the brief is weak.- Confirm one primary blog post type before drafting. Supported playbooks are
opinion / argument, explainer, comparison, product / deep dive, personal essay / rant, journal / dispatch, reflective / inspirational, and review.
- Preserve the core thesis all the way through the piece.
- Use section headings that move the argument forward, not generic SEO filler.
- Keep the article in paragraphs by default. Avoid bullet points and numbered lists unless the brief clearly requires them.
- Use
--- between major sections.
- Keep the voice closer to a sharp tech-magazine feature or analysis than to a generic SaaS explainer.
- Include
metadata.coverImage in every draft and published article.
- Include inline images in at least one substantive section, and follow
references/workflow.md for coverage on longer posts.
- Use a CTA only when it fits the archive's usual intensity.
Supporting Skills
Use the local shared review skills first because they are tuned for this repo's workflow:
../content-humanizer/SKILL.md../content-tone-review/SKILL.md../content-style-review/SKILL.md../fact-check/SKILL.md../source-url-check/SKILL.md../blog-post-ramsay-review/SKILL.md for blunt late-stage pressure testing../social-common/references/repetition-guard.md../blog-image-finder/SKILL.md../blog-image-uploader/SKILL.md when publishing image assets
If the session also includes external writing skills, use them in support of this skill instead of replacing it.
- Use
../copywriting/SKILL.md to sharpen structure or CTA options after the corpus voice is established.
- Use
$prompt-engineering-patterns principles to keep the workflow strict, progressive, and low-noise.
Keep the published corpus and shared repo references as the source of truth when those skills pull in a different style.
Output Shape
Unless the user asks differently, return:
- the final draft first
- the confirmed blog post type and any normalized playbook mapping
- a short note on the internal and external research that shaped the angle, including when no relevant internal examples existed
- the resolved blog source folder and publish output folder when they affected the work
- the key context files, corpus examples, and external sources used
- the selected cover image and inline image plan, plus any downloaded or uploaded asset paths when image work happened
- a short note on the outline or structural decisions that shaped the article
- a short note listing the voice cues you matched
- a
Review Ledger section with these lines in this order: Humanizer, Style review, Fact check, Source URL check, Affiliate compliance review when applicable, Tone review, Ramsay review
- on the
Ramsay review line, include the verdict, the X/15 score, and whether Must Fix is none, addressed, or waived by user
- a short workflow-completion note confirming that internal research, external research, outline validation, both writing passes, and the full review ledger all ran
- any assumptions or missing facts that could change the draft
- the saved draft path if you persisted the draft to the selected profile's blog draft root
- the One Horizon initiative title when a One Horizon record was created under
Ink - Blog
When invoked by a runner for an existing Blog initiative, return draft and review content compatible with the Blog Lifecycle Contract in ../one-horizon-context-setup/references/ink-initiative-hierarchy.md:
## Draft for the current article body## Review Notes for confirmed type, research summary, image notes, source notes, review ledger, and workflow summary## Automation for the runner checkpoint when the caller needs it
Files
- Read
references/corpus-spec.md for the blog archive schema.
- Read
references/draft-spec.md for unpublished draft storage.
- Read
references/workflow.md for the orchestration flow.
- Read
references/review-passes.md for pass order and responsibilities.
- Read
references/style-capture.md when extracting voice markers.
- Read
references/format-playbooks.md for article-shape rules.
- Read
references/affiliate-integration.md when partner, referral, commerce, or affiliate links are in scope.
- Read
../social-common/references/repetition-guard.md when comparing a draft against recent corpus patterns.
- Read the selected profile's
blogPublishingConfig for the active blog source and publish folders.
- If a One Horizon tool call fails, read
../one-horizon-context-setup/references/mcp-readiness.md for recovery.
- Resolve the active Ink profile with
../one-horizon-context-setup/references/ink-profile-contract.md, then resolve the author with One Horizon MCP tools and load relevant author-scoped context from the selected workspace using the naming and missing-doc contract in ../one-horizon-context-setup/references/context-doc-templates.md.
- Read
../../content/blog/README.md only if the selected profile's blogPublishingConfig points into this repo and the local naming is unclear.
- Read
../linkedin-social-writer/references/mcp-tools.md before using local research tools.
- Read
../blog-image-finder/references/setup.md before sourcing external blog images.
- Read
../blog-image-uploader/references/setup.md only when blog images should be uploaded to the blog image bucket.
- Use
assets/templates/blog-article.mdx when you need to persist an unpublished draft manually.
