GitHub

Cloud Rules

Audit rules powered by cloud services - what they check and what they cost

Cloud rules are regular audit rules that need data your machine doesn’t have - AI models, full filter lists, search demand data, or a shared link cache. They run automatically when you’re logged in and report skipped when you’re not.

The rules

RuleWhat it checksCost
ai/llm-parsabilityHow well LLMs can parse and understand each page1 credit per page
ai/page-type-matchAI-classified page type agrees with declared structured data1 credit per page (shared with parsability)
ai/site-typeClassifies the overall site type (blog, ecommerce, saas, docs, …)5 credits per audit
eeat/authority-signalsPer-page authorship, citations, and outbound references1 credit per page
adblock/element-hidingElements hidden by EasyList cosmetic rules5 credits per audit, shared
adblock/blocked-linksLinks and resources ad blockers would blockshared with above
adblock/privacy-blockedTrackers that EasyPrivacy filter lists would blockshared with above
links/dead-linksExternal links verified through a shared global cache1 credit per 100 URLs
gaps/keywordsKeywords competitors rank for that you don’t25 credits per audit, opt-in
gaps/contentTopic clusters with search demand you don’t cover25 credits per audit, opt-in

The three adblock rules share a single 5-credit blocklist check per audit - enabling all three costs the same as enabling one.

The two gap-analysis rules are disabled by default because of their cost. Enable them explicitly:

squirrel.toml
toml
[rules]
enable = ["gaps/keywords", "gaps/content"]

How they run

Cloud rules match rule filters like any other rule - enable = ["*"] (the default) includes them. During an audit, a dedicated cloud phase batches all the required service calls, then rules read the results. You’ll see the estimate in the start banner and the actual spend per service in the audit summary:

☁ Cloud: 67 credits spent (ai_parse 22, authority_signals 22, site_type 5, ...)

Skip behavior

A cloud rule never fails your audit. When it can’t run, it reports skipped with a reason:

Skip reasonMeaning
not-authenticatedNot logged in - run squirrel auth login
insufficient-creditsBalance too low - top up or lower the page count
credit-cap-reachedThe audit hit max_credits_per_audit; work was truncated at the cap
service-unavailableThe cloud service errored; charges for failed calls are refunded

Controlling spend

The per-audit cap, confirmation threshold, and --yes flag are covered in Credits & Pricing.

To opt out of specific cloud rules while keeping the rest:

squirrel.toml
toml
[rules]
disable = ["adblock/*", "links/dead-links"]

Or disable all cloud calls for a project:

squirrel.toml
toml
[cloud]
enabled = false

Type to search…

↑↓ navigate open esc close