charlieviettq/awesome-agent-skill / .claude/skills/algo-seo-backlink
"algo-seo-backlink"
Verified"Evaluate backlink quality using Domain Authority, Domain Rating, and trust metrics. Use this skill when the user needs to assess link profile health, identify toxic backlinks, or plan link building strategy — even if they say 'check my backlinks', 'link building', or 'domain authority analysis'.".
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.1 KB |
| examples/sample_scenario.md | markdown | 6.0 KB |
| references/competitor-gap.md | markdown | 8.3 KB |
| references/toxicity-scoring.md | markdown | 9.1 KB |
references/toxicity-scoring.md
View on GitHub# Toxicity Scoring Methodology
Toxicity scoring assigns a numerical risk score to each referring domain so you can prioritize disavow candidates objectively. The goal is to separate "weak but harmless" links from "actively harmful" links — disavowing the former wastes authority; ignoring the latter invites manual penalties.
---
## Scoring Model
Each referring domain receives a **Toxicity Score (TS)** from 0–100 computed from four signal groups. Higher score = more toxic.
```
TS = w_auth × S_auth + w_rel × S_rel + w_pattern × S_pattern + w_anchor × S_anchor
```
Default weights (sum to 1.0):
| Signal Group | Variable | Weight |
|---|---|---|
| Authority inversion | S_auth | 0.30 |
| Topical irrelevance | S_rel | 0.25 |
| Link pattern signals | S_pattern | 0.30 |
| Anchor text risk | S_anchor | 0.15 |
Weights are tunable. If you have a history of PBN attacks, raise `w_pattern`. If you're in a niche where off-topic links are common (e.g., press releases), lower `w_rel`.
---
## Signal Group Definitions
### S_auth — Authority Inversion Score (0–100)
A low-authority domain linking to a high-authority target is suspicious; the direction should normally be opposite.
```
S_auth = max(0, 100 − DR_linking) if DR_linking < 15
= max(0, 30 − DR_linking) if 15 ≤ DR_linking < 30
= 0 if DR_linking ≥ 30
```
Practical table:
| DR of linking domain | S_auth |
|---|---|
| 0–4 | 96–100 |
| 5–9 | 91–95 |
| 10–14 | 86–90 |
| 15–19 | 11–15 |
| 20–29 | 1–10 |
| 30+ | 0 |
**Note:** DR 0–4 sites that have existed < 30 days get a flat `S_auth = 100` regardless.
---
### S_rel — Topical Irrelevance Score (0–100)
Relevance is estimated from topic overlap between the linking page and the target site's category. Use the linking page's category tag from Ahrefs/Moz, or run a simple keyword cosine similarity if you have crawl access.
**Rule-based proxy** (when you lack ML-based topic classification):
1. Pull the top-3 topic categories of your target domain (e.g., "E-commerce > Electronics")
2. Pull the Ahrefs-reported category for the linking domain
3. Apply this scoring table:
| Overlap level | Definition | S_rel |
|---|---|---|
| Exact match | Same L2 category | 0 |
| Adjacent | Same L1, different L2 (e.g., both "E-commerce" but different sub) | 20 |
| Distant | Different L1 but plausible (e.g., "Tech" → "Electronics") | 50 |
| Unrelated | No thematic overlap (e.g., casino → software) | 80 |
| Adversarial | Adult/gambling/pharma → non-adult target | 100 |
Foreign-language mismatch that also has unrelated topic: add +15 (cap at 100).
---
### S_pattern — Link Pattern Score (0–100)
This signal group uses boolean indicators, each contributing a fixed score. Scores stack and are capped at 100.
| Pattern | Detection method | Score |
|---|---|---|
| Footer/sitewide link | Same anchor text on 100+ pages of same domain | +40 |
| PBN indicator | Ahrefs: referring domain has 0 organic traffic + DR < 20 + same IP C-block as ≥ 3 other linking domains | +50 |
| Link farm | Page linking to you also links to 50+ unrelated root domains on same page | +45 |
| Paid link signal | Anchor is exact-match keyword + linking page is clearly advertorial without disclosure | +30 |
| Age manipulation | Domain registered < 6 months ago with 50+ outbound links | +25 |
| Redirect chain | Link passes through ≥ 2 hops before reaching target | +10 |
| Reciprocal link | You also link back to this domain from a non-editorial context | +15 |
**PBN C-block check** (quick shell method):
```bash
# Get /24 block from IP
dig +short example.com | awk -F. '{print $1"."$2"."$3}'
# Compare across all suspicious domains in your export
```
---
### S_anchor — Anchor Text Risk Score (0–100)
Anchor text alone is rarely sufficient to trigger toxicity, but it amplifies risk from other signals.
```
S_anchor = base_anchor_risk × amplifier
```
**Base anchor risk:**
| Anchor type | S_anchor base |
|---|---|
| Branded (your exact brand name) | 0 |
| Naked URL | 5 |
| Generic ("click here", "website") | 10 |
| Partial-match keyword | 30 |
| Exact-match keyword | 55 |
| Exact-match + location ("buy widgets NYC") | 70 |
| Unrelated or spun text | 65 |
**Amplifier** (multiplicative, 1.0–1.5):
- Apply 1.3× if the same anchor is used on every page of the linking domain (sitewide)
- Apply 1.5× if anchor contains money keywords in a niche known for spam (CBD, casino, forex, pharma)
- Default amplifier: 1.0
Cap result at 100.
---
## Composite Score Interpretation
```
TS_final = 0.30 × S_auth + 0.25 × S_rel + 0.30 × S_pattern + 0.15 × S_anchor
```
| TS_final | Classification | Recommended action |
|---|---|---|
| 0–19 | Clean | No action |
| 20–39 | Weak | Monitor; no disavow |
| 40–59 | Suspicious | Manual review required |
| 60–79 | Likely toxic | Add to disavow candidate list |
| 80–100 | Confirmed toxic | Disavow immediately |
---
## Worked Example
**Scenario:** A domain `cheapmeds-discounts.ru` links to an e-commerce electronics site.
**Step 1 — S_auth**
DR = 3 → `S_auth = 97`
**Step 2 — S_rel**
Linking domain category: Pharma/Health. Target site: Electronics E-commerce.
Overlap: Adversarial (pharma → non-pharma) → `S_rel = 100`
**Step 3 — S_pattern**
- Footer/sitewide? Yes (same anchor on 200+ pages): +40
- PBN? 0 organic traffic + DR 3 + shares C-block with 4 other linking domains: +50
- Subtotal before cap: 90 → `S_pattern = 90`
**Step 4 — S_anchor**
Anchor: "cheap cialis online" → exact-match + pharma niche amplifier
Base: 70 (exact-match + location-style) × 1.5 = 105 → capped at 100 → `S_anchor = 100`
**Composite:**
```
TS = 0.30×97 + 0.25×100 + 0.30×90 + 0.15×100
= 29.1 + 25.0 + 27.0 + 15.0
= 96.1 → Confirmed toxic → Disavow immediately
```
---
## Batch Scoring Procedure
When processing a full backlink export (typical: hundreds to thousands of domains):
1. **Deduplicate first.** Collapse all links from the same root domain into one row. Use the highest-authority page on that domain as the representative link.
2. **Auto-score using rules.** Apply the four signal groups programmatically. Most tools export enough data to compute S_auth (DR), S_rel (category), and S_anchor (anchor text). S_pattern requires cross-domain analysis.
3. **Triage by bucket:**
- TS ≥ 80: auto-add to disavow file
- TS 60–79: export to manual review spreadsheet
- TS < 60: no action
4. **Manual review of the 60–79 bucket.** For each: open the linking page, check for editorial context, confirm pattern signals. Override up or down as needed.
5. **Never disavow the 40–59 range** without a strong editorial reason. These are usually old, low-quality directories or scrapers — mostly harmless.
---
## Disavow File Format
Google's disavow tool accepts domain-level and URL-level entries:
```
# Generated 2025-01-15 | Batch: 2025-Q1 audit
# TS ≥ 80: confirmed toxic
domain:cheapmeds-discounts.ru
domain:casino-backlinks-4u.net
domain:seo-links-farm-xyz.com
# TS 60–79: reviewed manually, confirmed
domain:irrelevant-directory-1998.info
```
**Rules:**
- Use `domain:` prefix to disavow all links from a root domain; use full URLs only when you want to preserve other links from the same domain (rare)
- One entry per line; `#` comments allowed
- Filename must be `.txt`, UTF-8 encoded
- Submit via Google Search Console → Links → Disavow
---
## Common False Positives
These patterns look toxic but usually aren't:
| Pattern | Why it looks toxic | Why it isn't |
|---|---|---|
| High-DA news site, exact-match anchor | S_anchor elevated | News sites editorially choose anchors; trust the editorial context |
| Old directory (DR 5, created 2003) | Low DR | Directories from pre-2010 often have historic trust Google discounts but doesn't penalize |
| Foreign-language domain, irrelevant topic | S_rel = 80 | If it's a legitimate import/export partner mention, relevance scoring fails |
| Competitor's site linking to yours | Pattern signals unclear | Competitor links are rarely manufactured; usually accidental or comparative |
When manual review reveals a false positive, override the classification and document the reason. Keep a suppression list to avoid re-flagging the same domains in future audits.
---
## Limitations of This Model
- **DA/DR are estimates.** Moz DA and Ahrefs DR disagree on 20–30% of domains. A domain scoring DR 12 in Ahrefs might be DA 35 in Moz. Use the tool you're most familiar with consistently across an audit, not a mix.
- **Google does not publish a toxicity API.** This model approximates Google's spam signals based on observed penalty patterns and Search Quality Evaluator Guidelines. It is not derived from Google's actual classifier.
- **Manual penalties ≠ algorithmic filtering.** Google's SpamBrain may already be ignoring a link algorithmically; a manual action is a separate, human-reviewed process. If you have no manual action in Search Console, your toxic links may already be discounted.
- **Disavow is irreversible in the short term.** Once submitted, a disavow file takes weeks to process and Google does not "un-process" it immediately if you remove entries. Err toward caution.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |