Borrowing it
Nothing to install: this file belongs to prashishh/seo-geo-report-engine. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/prashishh/seo-geo-report-engine/main/.agents/skills/keyword-research/SKILL.mdgit clone --depth 1 https://github.com/prashishh/seo-geo-report-engineWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/prashishh/seo-geo-report-engine/keyword-research)<a href="https://agentmods.dev/skills/prashishh/seo-geo-report-engine/keyword-research"><img src="https://agentmods.dev/badge/skills/prashishh/seo-geo-report-engine/keyword-research/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/prashishh/seo-geo-report-engine/keyword-research"><img src="https://agentmods.dev/badge/skills/prashishh/seo-geo-report-engine/keyword-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00100 | $0.01462 |
| Opus 5 | $0.00050 | $0.00731 |
| Sonnet 5 | $0.00020 | $0.00292 |
| Haiku 4.5 | $0.00010 | $0.00146 |
Grade A, and why
keyword-research scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
keyword-research
Turns a handful of seeds into a prioritized keyword map (clusters → target pages). The IP is
the scoring: don't just rank by volume — rank by winnable volume given the client's authority.
Prefer Ahrefs MCP (see knowledge/ahrefs-mcp-map.md); call doc on a tool before first use.
Methodology (PERCEIVE → ANALYZE → VALIDATE → ACT)
PERCEIVE — gather. Resolve the project (./bin/mkt config show --project <client>); read
client.yml for domain, locales, competitors, seeds. Establish the client's ceiling:
site-explorer-domain-rating (our DR) — this sets the KD bar we can realistically win.
- Expand each seed with
keywords-explorer-matching-terms(everything containing the seed),keywords-explorer-related-terms("also rank for"), andkeywords-explorer-search-suggestions(autocomplete long-tail). Pullkeywords-explorer-overviewfor volume, KD, CPC, parent topic. - Localize:
keywords-explorer-volume-by-countryfor each target locale — don't assume US volume. - Seasonality check on head terms:
keywords-explorer-volume-history(flag spiky vs evergreen).
ANALYZE — cluster + score. Two-axis clustering:
- Intent — informational / commercial / transactional / navigational. Infer from the term and
SERP (
serp-overviewon a sample to confirm — features, page types). Tag every keyword. - Topic — group by Ahrefs parent topic + shared head term so one cluster = one target page.
- Priority score per cluster:
score = volume × intent_weight × feasibility, whereintent_weight≈ transactional 1.0 / commercial 0.8 / informational 0.5 / navigational 0.2 (tune to the client's funnel), andfeasibility = 1if cluster median KD ≤ our DR-implied ceiling, scaling down as KD rises above it. Surface a few high-volume/low-feasibility terms as "later" so the client sees them, but rank winnable clusters first.
VALIDATE — every recommendation falsifiable. For each top cluster state:
- Observation — the volume/KD/intent data that motivates it (cite the Ahrefs tool + date).
- Dependency — what must be true to win it (e.g. "our DR ≥ median KD of SERP", "we have a page type matching the dominant intent", "topic is on-strategy for the ICP").
- How we'd know this failed — a leading indicator: e.g. "after publishing, page stuck below
position 20 at 8 weeks in
rank-tracker-overview" or "SERP is dominated by forums/UGC we can't displace." Note seasonal terms whose flat traffic is expected off-peak.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 89 lines · 100 tokens per session scan A 8488cf0a589f
keyword-research is a skill published in the GitHub repository prashishh/seo-geo-report-engine (5 stars, last pushed 2mo ago), licensed MIT. It adds 100 tokens to every session and 1,462 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
content-amplifier
Use when the user asks to "amplify influencer content with paid media", "set up whitelisting or Spark Ads", "decide which posts to boost", "repurpose influencer content", "turn one video into multiple ads", or "build a UGC asset library"; produces (paid mode) a content-selection scorecard, a paid amplification…
email-sequence-designer
Use when the user asks to "design a welcome flow", "set up an abandoned-cart sequence", "build a light re-engagement branch inside a lifecycle flow", or "plan a cold-outbound sequence"; produces general lifecycle automation flows (welcome, cart, browse-abandon, post-purchase, in-flow re-engagement, B2B cold outbound)…
reactivation-specialist
Use when the user asks to "build a win-back campaign", "re-engage lapsed subscribers", "run a re-permission / re-consent sweep", or "sunset my dead list"; produces a closed-loop reactivation program — a lapsed-cohort definition, a staged offer ladder, a re-consent (re-permission) capture step, and a sunset-confirm /…
campaign-planner
Use when the user asks to "plan an influencer campaign", "build a campaign blueprint", "track or close a creator campaign", or "record a late campaign correction"; produces the plan and, when requested, a non-canonical evidence tracker with scoped identity, publication, reconciliation, close, and reopen receipts. Not…
product-feed-optimizer
Use when the user asks to "optimize my Shopping feed", "fix product disapprovals", "improve product titles/attributes", or "build feed-driven PMax asset groups"; audits and rewrites the Shopping/Performance Max product feed — title/description patterns, required and recommended attributes, GTIN/availability/price…
cold-outbound-sequencer
Use when the user asks to "build a B2B cold-outbound sequence", "design reply-triage branching", "plan a domain warmup / sending throttle", or "make my outbound CAN-SPAM / opt-in compliant"; produces a multi-step outbound sequence with reply-triage branches (positive / objection / referral / not-now / opt-out), a…