More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Editorial QA: Pre-Publish Review

    1

    A seven-gate pre-publish QA system: brief adherence and fact accuracy halt a piece, structure, AI patterns, and voice get flagged, and search and schema fixes ship in place.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Editorial QA: Pre-Publish Review

    Editorial QA: Pre-Publish Review

    Example session with this skill installed

    Run your QA pass on this draft before it publishes tomorrow. Brief: a beginner how-to targeting the query "how to write a changelog," with one required link to our release-notes guide. Draft excerpt: "In today's fast-paced software world, changelogs matter more than ever. Studies show that 68% of users read release notes before updating, and as one engineering leader puts it, 'a good changelog is a love letter to your users.' Beyond that, changelogs are important for many reasons, including transparency, trust, and more. Start by listing everything that changed in the release…"

    • Read your context and instructions
    • Compiled the editorial qa

    QA REPORT — "How to Write a Changelog" draft

    Gate 1 Brief adherence: 1 — It is a how-to and the target query is in the title, but the required link to the release-notes guide is missing, and the opening addresses nobody in particular instead of the beginner the brief names. Fixable in the edit; noted, not a halt on its own.

    Gate 2 Fact accuracy: 0 — HALT. "Studies show that 68% of users read release notes before updating" cites no study. No source was supplied with the draft, and the figure cannot be traced. The quote "a good changelog is a love letter to your users" is attributed to "one engineering leader" — no name, no source. Both are removed or sourced before this ships; an invented statistic and an anonymous quote are exactly the two classes this gate exists to stop.

    Gate 3 Structure: 1 — Steps begin reasonably, but the first 200 words are preamble. The how-to itself should open the piece.

    Gate 4 AI patterns: 1 — "In today's fast-paced software world" is a throat-clearing opener, and "including transparency, trust, and more" is sentence-end filler plus a list trailing off instead of landing. Both rewritten in the edit.

    Gates 5–7: voice is on-brand once the openers are cut (2); meta description missing — auto-fixed; no schema issues (2).

    VERDICT: Halt until the statistic and quote are sourced or deleted. Everything else is a same-pass fix. Reviewer time: about 15 minutes, because the halt fired at Gate 2 before any line editing began.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most content QA is either a skim that ships hallucinations or a 47-item checklist nobody completes honestly. This skill replaces both with a gate sequence ordered by economics: the cheapest checks that can stop a piece run first, so you never spend an hour polishing the voice of a draft that missed its brief or invented its statistics. You get the seven gates (brief adherence, fact accuracy, structure, AI-pattern audit, voice consistency, search compliance, links and schema), a disposition for every finding (halt, flag, or auto-fix), a 0–2 scorecard for each gate with a hard verdict rule, and a sampling protocol for programmatic sets where full audits are impossible — including the failure threshold that halts generation. It also covers the economics of the checklist itself: every check must name the failure class it catches and the last time it caught something, or it gets cut. Includes a fill-in QA report template, a pre-publish checklist, a complete worked example, escalation rules for turning repeat piece failures into process fixes, and a failure-mode list drawn from real QA breakdowns (zero-catch quarters, reviewer burnout, cherry-picked sampling).

    • It checks facts against sources you can supply or that can be verified in the session; a claim whose source does not exist will be flagged for removal, not magically verified.
    • Gate 1 is only as strong as the brief. If no brief exists, the skill reconstructs a minimal one and marks the gate as such — it cannot enforce a contract nobody wrote.
    • It judges editorial quality, not legal, regulatory, or safety compliance; regulated content still needs its own review.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    The gates are ordered so the cheap ones fire first — brief and fact halts take minutes and stop the pieces that would have eaten the most editing time. Most clean pieces clear in one pass, and the scorecard is a one-page record, not a form per paragraph.

    No. It gives an editor (human or agent) a fixed sequence, explicit halt conditions, and a scoring record so two reviewers reach the same verdict on the same draft. Judgment calls at the flag gates still belong to the editor.

    Start with the three-line reconstruction the skill describes (audience, purpose, target query) for one content stream, run the gates for a month, and let the catch log show you which brief fields actually matter for your content.

    New in Writing & Content