Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add AntonioBlago/peec-ai-skills --skill peec-content-intelgit clone --depth 1 https://github.com/AntonioBlago/peec-ai-skillsWrote 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/antonioblago/peec-ai-skills/peec-content-intel)<a href="https://agentmods.dev/skills/antonioblago/peec-ai-skills/peec-content-intel"><img src="https://agentmods.dev/badge/skills/antonioblago/peec-ai-skills/peec-content-intel/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/antonioblago/peec-ai-skills/peec-content-intel"><img src="https://agentmods.dev/badge/skills/antonioblago/peec-ai-skills/peec-content-intel.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 | $0.00117 | $0.03087 |
| Opus 5.5 | $0.00047 | $0.01235 |
| Sonnet 5.5 | $0.00023 | $0.00617 |
| Haiku 4.5 | $0.00012 | $0.00309 |
Grade A, and why
peec-content-intel 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 2mo 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 — 285 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Peec Content Intel
Role
For one Peec prompt that the brand is losing, produce one publish-ready content brief: sub-queries, verbatim buyer pains, competitor breakdown, outline, focus keywords, and outreach targets.
Input
project_id— Peec project (read fromsetup_state.jsonper pre-flight; do not re-resolve)prompt_id— the target prompt (must have ≥24h of data)- optional
date_range— default last 28 days language— read fromsetup_state.json(prompt_language); user can override per-run, but never default toensilentlytarget_country— read fromsetup_state.json; drives forum source picks (DE→reddit/r/de+gutefrage+t3n, AT→reddit+derstandard, US→reddit+quora, CH→reddit/r/de+r/fr) and SERP/GSC market filterspage_type— required. Must be one of the values insetup_state.page_type_taxonomy. If the user doesn't pass it, the skill infers the best fit from the funnel stage of the target prompt + competitor top-url classifications, then ASKS the user to confirm ("Suggested page_type:landing_page(Decision-stage, competitors use PRODUCT_PAGE). OK? [y/n/override]"). Never guess silently — a brief with the wrong page_type is a wasted publish cycle.business_type— read fromsetup_state.json. Used to validatepage_typeselection against the allowed taxonomy matrix.audience— read fromsetup_state.json.audience.primaryandaudience.pain_pointsfeed the brief's "Why this page wins" + "Voice / tonality" sections directly; do not restate them in the brief prose, USE them.
Output
One markdown brief per prompt, saved at briefs/<YYYY-MM-DD>_<prompt-slug>/brief.md, plus the raw data next to it (competitor-urls.json, forum-pains.json, scoring.json). No dashboards.
Every brief starts with a front-matter block that downstream skills (peec-report, peec-learn) consume for attribution:
---
brief_id: <YYYY-MM-DD>_<prompt-slug>
prompt_id: pr_xxxxxx
business_type: b2b-service
page_type: landing_page # MUST be in setup_state.page_type_taxonomy
target_url: /seo-retainer # planned publish path
funnel_stage: Decision # from the prompt's topic
audience:
primary: "Shop-Owner DACH, 3-20 MA, Shopify"
pain_hook: "3 Agenturen gewechselt, keine Ergebnisse"
success_metric_4w:
prompt_visibility: "0% → ≥15%"
zone_visibility: "4% → ≥20%"
---
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.
- 2mo ago First seen · 285 lines · 117 tokens per session scan A 6489b18f95e7
peec-content-intel is a skill published in the GitHub repository AntonioBlago/peec-ai-skills (10 stars, last pushed 6mo ago), licensed MIT. It adds 117 tokens to every session and 3,087 once invoked, about $0.0005 per session on Opus 5.5. 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
seo-content-blog
Spec, audit, competitor benchmark and build for blog articles and the blog template, so a post ranks on Google and gets cited by AI engines (ChatGPT, Perplexity, AI Overviews). Input: a keyword, topic, draft or post URL. Phase 1 spec: BLOG rows on top of the common page spec, covering the 12-element answer-first…
seo-content-collection-page
Spec, audit, competitor benchmark and build for e-commerce collection, category and product listing pages (PLPs) on Shopify, WooCommerce, Magento, BigCommerce, PrestaShop or a custom store. Runs four phases: the collection spec (COL-01 to COL-24 plus the common rows), an audit of the live page with the bundled…
seo-content-comparison-page
Spec, audit, competitor benchmark and build for the bottom-funnel pages that capture buyers at the decision moment: 'X vs Y', 'alternatives to X', 'best X for Y', the comparison hub, and audience or segment pages ('for agencies', 'for freelancers'). Runs four phases: the CMP spec with stable IDs on top of the common…
seo-content-product-page
Spec, audit, competitor benchmark and build for e-commerce product pages (PDPs), so they rank in Google and get recommended by the AI assistants that now link products directly. Input: a product page URL or a product brief (Shopify, WooCommerce, Magento, BigCommerce, PrestaShop, Wix, Webflow or custom). Runs four…
seo-content-service-page
Spec, audit, competitor benchmark and build for service pages and service landing pages (agencies, consultants, coaches, trades, clinics, law firms, SaaS), plus region and city pages. Input: a service, the business, a target area, the page URL if any. Runs four phases: the spec (SVC-01 to SVC-24 on top of the common C…
seo-local
Run spec, audit, competitor benchmark and build for local visibility, on the two assets a local business owns: the location page and the Google Business Profile, plus reviews and NAP citations. Spec: LOC, GBP, REV and CIT requirements with IDs. Audit: seoaudit.py and sectionaudit.py --type location on the page, a…