# Review Google Docs Article > Reviews a TrueProfit Google Docs article before publishing, one QA step at a time. Use this whenever the user wants to "review an article", "review this doc", "QA a blog doc", "check the links in this Google Doc", "run a link review", "find internal link opportunities", "check external links / references", or asks whether a doc's links and citations are ready to publish. Also trigger when the user pastes a Google Docs URL plus a list of "URLs to check" and asks what to add, remove, or fix. It also handles the CTA image (`[cta]` marker), "Further Reading" placement, and FAQ-redundancy checks. Steps so far: (1) Internal & External Links, (2) CTA image, (3) Further Reading placement, (4) FAQ redundancy — more to be added later. Prefer this skill over eyeballing the doc or checking links, the CTA, Further Reading, or the FAQ by hand. - Skill: `khanhvo2172-dotcom/review-google-docs-article` (Agent Skill, multi-file: 9 files) - Install (CLI): `npx skillmds@latest add khanhvo2172-dotcom/review-google-docs-article` - Raw SKILL.md: https://api.skillmd.com/api/skills/khanhvo2172-dotcom/review-google-docs-article/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: khanhvo2172-dotcom (https://skillmd.com/u/khanhvo2172-dotcom) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/khanhvo2172-dotcom/review-google-docs-article --- # Review a TrueProfit Google Docs Article This skill runs a structured pre-publish QA review of a TrueProfit blog article that lives in a Google Doc. The review is deliberately **modular** — each step is a self-contained QA pass with its own reference file. Steps 1–4 (links, CTA image, Further Reading, FAQ redundancy) exist today; future steps (tone/humanization, data benchmarks, formatting, …) slot in the same way. ## How to run a review 1. Ask which step(s) the user wants if they didn't say. Default to running **all available steps in order** (1 → 2 → 3 → 4). 2. For each requested step, open its reference file and follow it exactly. 3. Deliver each step's findings as a **report in chat** (not a file) using the structure the step defines. If running multiple steps, give one clearly-headed section per step. ## Available steps | # | Step | Reference file | What it checks | |---|------|----------------|----------------| | 1 | Internal & External Links | `references/step-1-internal-external-links.md` | Missing internal-link opportunities (with anchor-text proposals), duplicate links to remove, and external-reference accuracy (benchmark matching + DR≥65 authority + broken links) | | 2 | CTA image | `references/step-2-3-structure-checks.md` | Whether a `[cta]` marker qualifies for the CTA image (> 5 H2, Quick Recap excluded, FAQ counted) and, if so, **inserts** the CTA image trigger; warns on short articles | | 3 | Further Reading placement | `references/step-2-3-structure-checks.md` | Warns when a "Further Reading" block sits directly above the 2nd–5th H2 (report-only) | | 4 | FAQ redundancy | `references/step-4-faq-redundancy.md` | Flags FAQ questions that duplicate a heading's intent or are already answered in the body (report-only, model-driven) | Steps 2 and 3 share one pass over the doc's headings, so `scripts/review_structure.py` runs both together. Step 4 is model-driven (judgment about intent), no script. > **Adding a step later:** write `references/step-N-.md` describing that pass, then add > a row to the table above and a line to the "How to run" default order. Keep each step's > heavy detail in its own reference file so this orchestrator stays short. ## Inputs the user must provide - **Google Doc URL** — the article to review. **The URL's `tab=t.xxxx` parameter names the exact tab to review; that tab is the whole scope.** The doc must be readable by the link-checker app's Google service account, or shared as "anyone with the link can view". - **URLs to check** — the candidate internal links, one per line, in the format `https://www.trueprofit.io/blog/some-page | Page title`. The title after `|` is optional but helps judge relevance. This is the list Step 1 compares against the doc. - **Must-have URLs** *(Step 1, optional)* — a subset of links that *must* be present. Paste into the app's `Must-have URLs` box; matching rows come back flagged `⭐ Must-have` and are treated as must-insert (not subject to the selective judgment). - **Main keyword** *(Step 2 only)* — a kebab-case slug for the CTA link's `utm_campaign`. Don't ask up front: derive it from the doc and **confirm with the user** (see Step 2). If a step's required input is missing, ask for it before running that step. ## Ground rules for every step - **Review ONLY the tab in the URL — never the whole document.** A TrueProfit blog doc almost always has **sibling tabs** beside the article (an `n8n flow` draft, an `Insights & outline` notes tab, AI-research dumps). These are separate tabs, **not** scratch content inside the article, and must never be flagged as "delete these blocks." The Drive reader (`read_file_content`) concatenates *every tab* into one blob — do **not** use it for this skill. Read article text with **`scripts/dump_tab.py`** (it honors the URL's `tab=` param and emits just that tab, with links inline as `[anchor](url)`). The structure script (`review_structure.py`) is already tab-scoped the same way. If a `tab=` param is missing, the scripts fall back to the first tab and say so — confirm the intended tab with the user rather than reviewing everything. - **Advisory by default; one exception.** Every step produces *recommendations* the user applies themselves — **except Step 2's CTA insertion**, which edits the Doc in place via `scripts/review_structure.py`. Always dry-run and show the plan first; only `--apply` after the user confirms (especially the `utm_campaign` keyword). Never claim an edit was made unless the apply run actually succeeded. - **Be selective, not exhaustive.** For opportunities (like missing links), recommend only what genuinely makes the content more valuable or better for SEO. Quality over coverage. - **Show your reasoning.** For every recommendation, say *why* (SEO value, relevance, accuracy, authority) so the user can trust or overrule it quickly.