What is a Claude content audit skill?

Quick Answer: A Claude content audit skill is a custom instruction set you build inside Claude's agent framework that automates content analysis tasks, including traffic decay detection, keyword cannibalism flags, and refresh prioritisation. Once built, it runs repeatable audits in minutes instead of days, without a spreadsheet marathon.

Running a content audit manually is one of the most time-consuming tasks in B2B SaaS marketing. You pull data from Google Search Console, cross-reference it against your CMS, check for overlapping keywords across posts, and then try to make sense of what to fix first. Most teams do it quarterly at best, and even then, it's incomplete.

Claude Skills change that. By building a dedicated content audit skill, you give Claude a structured set of instructions it follows every time you run an audit. The output is consistent, repeatable, and fast enough that you can run it monthly without burning a full day.

This tutorial walks you through exactly how to build that skill, what to include in it, and how to get it flagging the three things that matter most: content decay, keyword cannibalism, and refresh priority.

What Is a Claude Content Audit Skill?

A Claude content audit skill is a saved, reusable instruction set that tells Claude how to behave when you hand it content data. Think of it as a standing operating procedure your AI analyst follows without you having to re-explain the brief every time.

Skills live inside Claude's agent environment and can reference uploaded files, structured prompts, and multi-step workflows. When you build one well, Claude acts like a content strategist who already knows your criteria, your priorities, and your definitions before you say a word.

For B2B SaaS teams, this matters because your content library grows fast, your ICP evolves, and the gap between what ranks and what converts widens over time. A skill that catches that drift early is worth more than a quarterly audit that's already six weeks out of date by the time anyone reads it.

What Your Content Audit Skill Should Cover

Before you write a single line of your skill file, decide what outputs you actually need. The three highest-value outputs for most B2B SaaS content teams are:

• Decay detection: Which posts have lost 20%+ of impressions or clicks over the past 90 days
• Cannibalism flags: Which URLs are competing for the same keyword, splitting ranking signals
• Refresh recommendations: Which posts are worth updating versus retiring, based on traffic potential and content quality signals

Build your skill around these three outputs and everything else becomes supporting logic.

How to Build the Skill: Step by Step

Step 1: Prepare Your Data Inputs

Claude cannot pull live data from Google Search Console or your CMS on its own. You need to export it and feed it in. For each audit run, prepare:

• A GSC export covering the last 90 days (queries, URLs, impressions, clicks, average position)
• A CMS export with URL, publish date, last updated date, word count, and target keyword
• Optionally, a Screaming Frog or Ahrefs crawl export for internal link data

Save these as CSV files. Claude handles CSV well and can cross-reference across multiple files in a single session.

Step 2: Write Your SKILL.md File

The SKILL.md file is the core instruction document Claude reads before doing anything. It sets the rules, definitions, and output format for every audit run.

Here is a working structure for your content audit SKILL.md:

# Content Audit Skill - SaaS Hackers Framework

## Purpose
You are a content strategist auditing a B2B SaaS blog. Your job is to analyse the uploaded data files and produce a structured audit report covering decay detection, keyword cannibalism, and refresh recommendations.

## Definitions
- Decay: Any URL that has lost 20% or more in clicks or impressions compared to the prior 90-day period
- Cannibalism: Two or more URLs ranking in positions 1-20 for the same primary keyword with fewer than 3 positions separating them
- Refresh Priority Score: A 1-5 score based on traffic potential (current impressions), content age (days since last update), and position (closer to page 1 = higher priority)

## Inputs
You will receive up to three CSV files:
1. gsc_data.csv - Google Search Console export
2. cms_data.csv - CMS content inventory
3. crawl_data.csv - Optional crawl export

## Output Format
Produce three sections:
1. Decay Report (table: URL, current clicks, prior clicks, % change, decay flag)
2. Cannibalism Report (table: keyword, competing URLs, positions, recommendation)
3. Refresh Priority List (table: URL, priority score, reason, recommended action)

## Rules
- Flag decay only where the drop is sustained across at least 4 weeks, not a single bad week
- For cannibalism, always recommend which URL to keep as the primary and which to consolidate or redirect
- Refresh Priority Score 4-5 = act this month. Score 1-2 = monitor or retire.
- Do not recommend retiring a post unless it has fewer than 50 clicks in 90 days AND sits below position 30 for all tracked queries

This instruction set is specific enough that Claude produces consistent outputs across different team members running the same audit.

Step 3: Add Concrete Examples to the Skill File

The single biggest improvement you can make to any Claude skill is adding worked examples. Prose instructions alone leave room for interpretation. A short example closes that gap.

Add a section to your SKILL.md like this:

## Example Output (Decay Report)

| URL | Current Clicks | Prior Clicks | % Change | Flag |
|---|---|---|---|---|
| /blog/saas-pricing-models | 420 | 610 | -31% | DECAY |
| /blog/customer-onboarding | 880 | 900 | -2% | OK |

## Example Cannibalism Flag

Keyword: "saas churn reduction"
Competing URLs: /blog/reduce-saas-churn (pos 8), /blog/churn-rate-saas (pos 11)
Recommendation: Consolidate /blog/churn-rate-saas into /blog/reduce-saas-churn. Set 301 redirect.

Claude reads these examples and mirrors the format exactly. Your output arrives ready to paste into a report or share with a client.

Step 4: Upload the Skill and Test It

In Claude's interface, create a new Project and attach your SKILL.md as a project instruction file. This means Claude reads it at the start of every conversation in that project without you pasting it manually.

Run your first test with a small data sample, around 20-30 URLs, before committing your full content library. Check that:

• Decay flags match what you would flag manually
• Cannibalism detection is catching real overlaps, not false positives from loosely related keywords
• Refresh Priority Scores feel calibrated to your actual content situation

Adjust your definitions in the SKILL.md if the first run produces too many false positives. Tightening the decay threshold from 20% to 25%, for example, removes noise without losing the genuine problems.

Step 5: Build a Repeatable Monthly Workflow

The skill is only as useful as the cadence you run it on. Set a fixed monthly date, export your GSC and CMS data on the same day each month, and run the audit in the same Claude Project. This gives you comparable outputs month-over-month so you can track whether your refresh actions are working.

A monthly cadence is achievable for most B2B SaaS teams because the skill does the analysis. Your job becomes reviewing the output and making decisions, not building the audit from scratch.

How to Handle the Three Core Outputs

Reading the Decay Report

Decay is not always a problem worth fixing. A post that dropped 25% in clicks but still sits at position 4 for a high-volume keyword is a different situation from a post that dropped 25% and is now on page 3.

Use the decay report to prioritise by traffic opportunity, not just by percentage drop. Posts with high impressions and declining clicks are your highest-priority targets because they are already visible but losing ground. If you want a wider framework for fitting this into your overall acquisition plan, the B2B SaaS digital marketing guide is a useful reference point.

Acting on Cannibalism Flags

Keyword cannibalism is one of the most common structural problems in B2B SaaS content libraries that have been publishing for two or more years. Two posts competing for the same keyword split your link equity and confuse Google about which URL to rank.

When Claude flags cannibalism, the decision tree is straightforward:

1. Which URL has more backlinks? Keep that one as primary.
2. Which URL is more comprehensive and better matches search intent? If different from step 1, consider a rewrite and consolidation.
3. Redirect the secondary URL to the primary. Update internal links to point to the primary only.

Do not delete the secondary URL without setting a 301 redirect. You lose the link equity.

Using the Refresh Priority List

A Refresh Priority Score of 4 or 5 means the post has real traffic potential (high impressions), is losing ground (dropping position), and has not been updated recently. These posts respond well to refresh because Google is already indexing them and a quality update can push them back up within 4-8 weeks.

Score 1-2 posts are a different call. If the post has under 50 clicks in 90 days and ranks below position 30 for everything, the honest question is whether the topic is worth owning at all. Sometimes retiring and redirecting is the right move. Sometimes the post covers a topic your ICP genuinely needs, and a full rewrite is justified despite low current traffic.

Claude will not make that call for you, but the data it surfaces makes the decision much faster.

What Makes This Skill Better Than a Generic ChatGPT Prompt

The difference between a one-off prompt and a Claude skill is repeatability and specificity. A prompt you write fresh each time drifts. You phrase things differently, forget to include a definition, or skip a section when you're in a hurry.

A skill file is fixed. It contains your exact definitions, your output format, and your decision rules. Every audit run produces comparable output because the instructions do not change. For B2B SaaS teams managing content across multiple products, personas, or markets, that consistency is what makes the audit actually useful over time.

Nicholas Porter's Medium audit post, which covered a Claude CLI skills audit, found that leading with the single most important instruction and including concrete code or format examples were the two factors that most improved output quality. The SKILL.md structure above applies both of those lessons.

FAQs

Q: Can Claude pull data directly from Google Search Console for a content audit?

A: Claude does not have a native GSC integration. You export your data as a CSV from GSC and upload it to the Claude Project session. Claude then analyses the exported data according to your skill instructions. Some third-party MCP connectors are emerging that can bridge this gap, but a CSV export is the most reliable method for most teams today.

Q: How many URLs can Claude handle in a single content audit session?

A: Claude's context window handles several hundred rows of CSV data comfortably in a single session. For content libraries above 500 URLs, split the audit into segments by content category or publication date and run separate sessions. The skill instructions stay the same across sessions, so the outputs remain consistent.

Q: How is a Claude content audit skill different from an SEO tool like Ahrefs or Semrush?

A: SEO tools give you data. A Claude content audit skill gives you decisions. Ahrefs shows you which posts are declining. The Claude skill takes that data, applies your specific definitions and thresholds, cross-references it against your CMS inventory, and produces a prioritised action list with recommendations already written. The two work best together: pull the data from your SEO tool, feed it into the Claude skill, and get the analysis done in one session. If you're reviewing outside support for that broader search strategy, this list of B2B SaaS SEO agencies may help.

Q: How often should I run a content audit using this skill?

A: Monthly is the right cadence for most B2B SaaS teams publishing 4 or more posts per month. Quarterly audits miss decay that compounds over time. Monthly audits with a consistent skill output let you catch drops early, act on them within the same quarter, and measure whether your refresh actions are working before the next audit cycle.

Q: Is SaaS Hackers' approach to Claude skills relevant for small content teams?

A: The approach works for any team size. A solo content marketer running monthly audits with this skill saves 3-4 hours per audit cycle compared to doing it manually in a spreadsheet. Larger teams benefit from the consistency: every team member running the audit produces the same output format, which makes review and decision-making faster across the board.