CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

jbaruch/blog-writer

Write and revise developer blog posts or run a humanizer / humanization pass on existing drafts using the AI anti-pattern catalog, structural audits, and reusable personal and explicitly selected corporate writing identities.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

scoped-edit.mdskills/blog-writer/references/

Scoped prose work

Use this route for an existing artifact's humanization, review, copyedit, or local revision, and for a non-blog writing assignment. A conference abstract, email, or technical note keeps its own format and purpose. New blog creation uses the main skill's full workflow. An existing blog's limited editing request uses this route too.

Establish the assignment

Read the supplied prose, current conversational framing, and requested scope. For writing from notes, use those notes as the source. Ask only for information that changes the work and cannot be inferred; do not require an interview, blog home, slug, CTA, outline approval, identity creation, or shape-history record. A review-only request returns findings and suggestions without changing the supplied artifact.

Use explicitly supplied identities or an existing project selection when available. Read their relevant voice and editorial resources; do not discover or activate a corporate identity from the topic or author. When there is no selected personal identity, formal personal voice calibration is not applicable; when one is selected but its same-mode passages are missing, it is unresolved. Neither state prevents using assignment evidence. If a requested identity is unavailable, disclose that limitation rather than silently substituting another. Ask for a replacement only when the task needs it to proceed.

Read skills/blog-writer/references/voice-calibration.md and keep its short preservation contract in working notes. Distinguish the author's framing, endorsed collaborative copy, and unapproved assistant wording. Keep the original and pre-edit prose for comparison. Do not create research-bank or publication artifacts for a scoped edit.

Perform the requested work

Read skills/blog-writer/references/ai-anti-patterns.md. For a humanization pass, run the three-pass review in skills/blog-writer/references/process.md Phase 3, starting at "Pass 1" and ending with the proportionality check. These are editorial checks, not a request to run the surrounding blog intake, drafting, or publication phases. For a narrower copyedit or local correction, inspect the changed passages and their context under the same catalog boundaries; keep unaffected wording. Do not broaden an edit just to remove every stylistic scanner match elsewhere in the artifact.

Scoped work has no outline phase, so audit 1 (theme explicitness) in skills/blog-writer/references/structural-audits.md never runs unless it runs here. Run it on the prose: mark every sentence that states the artifact's point rather than showing it, and keep one. A sentence that names the theme after the material already implied it is the most common survivor of a sentence-level pass, because no numbered pattern owns it.

Execute skills/blog-writer/references/sweep-review.md for mechanical results, contextual retention, and the final artifact check. Treat the raw sweep independently from the editorial judgment. Preserve neutral exposition and supported attitude alike. Correct or flag unsupported facts; never invent a joke, experience, or source to make prose sound human.

Apply any explicitly selected corporate editorial review within its scope. Then verify source-to-revision preservation and within/between-paragraph argument flow. Repair any regression before describing the work as ready, including changes introduced by another review pass. Avoid repeated rewrites whose sole purpose is to change sentence counts.

Return the result

Return the requested draft, edit, or review in the requested medium. For review-only work, report proposed changes and unresolved findings; do not write them into the source file. For edits, briefly identify consequential changes and where the protected functions survive. Use concrete source/revision passages, not an assurance that the voice survived. Report the independent states from the voice-calibration reference, including retained scanner hits and unresolved evidence. Combine routine successes concisely; expand problems and material decisions. Finish here without blog-shape recording or publication bookkeeping.

README.md

tile.json