This is a library of rules for navigating through common consent popups on the web. These rules can be run in a Chrome extension, or in a Playwright-orchestrated headless browser. Using these rules, opt-in and opt-out options can be selected automatically, without requiring user-input.
The web extension can be built with the following steps:
# Download dependencies
npm ci
# Build the extension
npm run prepublish
The extension-specific code can be found in the addon
directory. There are two versions of the
addon (found under dist/addon
after building), one for mv3
version for Chromium-based browsers, and a firefox
version for Firefox.
You can load these in Chrome
in developer mode, and in Firefox
as a temporary addon.
The library's functionality is implemented as a set of rules that define how to manage consent on a subset of sites. These generally correspond to specific Consent Management Providers (CMPs) that are installed on multiple sites. Each CMP ruleset defines:
- If the site is using that CMP
- If a popup is displayed
- Steps to specify an 'opt-in' or 'opt-out' consent for the CMP.
- Optionally, a test if the consent was correctly applied.
There are currently three ways of implementing a CMP:
- As a JSON ruleset, intepreted by the
AutoConsent
class. - As a class implementing the
AutoCMP
interface. This enables more complex logic than the linear AutoConsent rulesets allow. - As a Consent-O-Matic rule. The
ConsentOMaticCMP
class implements compability with rules written for the Consent-O-Matic extension.
Sometimes the opt-out process requires actions that span across multiple pages or iframes. In this case it is necessary to define stages (each corresponding to a separate page context) as separate rulesets. Each one, except the very last stage, must be marked as intermediate using the intermediate: true
flag. If the intermediate
flag is not set correctly, autoconsent may report a successful opt-out even if it is not yet finished.
By default, rules will be executed in all top-level documents. Some rules are designed for specific contexts (e.g. only nested iframes, or only specific URLs). This can be configured in runContext
field (see the syntax reference below).
An autoconsent CMP rule can be written as either:
- a JSON file adhering to the
AutoConsentCMPRule
type. - a class implementing the
AutoCMP
interface, or- common JSON rules are available as reusable functions in dom-actions.ts. You can also use existing class-based rules as reference.
In most cases the JSON syntax should be sufficient, unless some complex non-linear logic is required, in which case a class is required.
Both JSON and class implementations have the following components:
name
- to identify this CMP.detectCMP
- which determines if this CMP is included on the page.detectPopup
- which determines if a popup is being shown by the CMP.optOut
- a list of actions to do an 'opt-out' from the popup screen. i.e. denying all consents possible.optIn
- a list of actions for an 'opt-in' from the popup screen.- (optional)
prehideSelectors
- a list of CSS selectors to "pre-hide" early before detecting a CMP. This helps against flickering. Pre-hiding is done using CSSopacity
andz-index
, so be it should be used with care to prevent conflicts with the opt-out process. - (optional)
intermediate
- a boolean flag indicating that the ruleset is part of a multi-stage process, see the Intermediate rules section. This isfalse
by default. - (optional)
runContext
- an object describing when this rule should be tried:main
- boolean, set totrue
if the rule should be executed in top-level documents (default:true
)frame
- boolean, set totrue
if the rule should be executed in nested frames (default:false
)urlPattern
- string, specifies a regular expression that should match the page URL (default: empty)
- (optional)
test
- a list of actions to verify a successful opt-out. This is currently only used in Playwright tests.
detectCMP
, detectPopup
, optOut
, optIn
, and test
are defined as a set of checks or actions on the page. In the JSON syntax this is a list of AutoConsentRuleStep
objects. For detect
checks, we return true for the check if all steps return true. For opt in and out, we execute actions in order, exiting if one fails. The following checks/actions are supported:
Many rules use ElementSelector
to locate elements in a page. ElementSelector
can be a string, or array of strings, which are used to locate elements as follows:
- By default, strings are treated as CSS Selectors via the
querySelector
API. e.g.#reject-cookies
to find an element whoseid
is 'reject-cookies'. - Strings prefixed with
xpath/
are Xpath selectors which can locate elements in the page viadocument.evaluate
. e.g.xpath///*[@id="reject-cookies"]
can find an element whoseid
is 'reject-cookies'. - If an array of strings is given, the selectors are applied in array order, with the search scope constrained each time but the first match of the previous selector. e.g.
['#reject-cookies', 'button']
first looks for an element withid="reject-cookies"
, then looks for a match forbutton
that is a descendant of that element. If one of the selectors returns an element that has ashadowRoot
property, the next selector will run within that element's shadow DOM. This is the main difference from nested CSS selectors, which do not cross shadow DOM boundaries.
For example, consider the following DOM fragment:
<open-shadow-root-element>
<button>X</button>
</open-shadow-root-element>
Then ['open-shadow-root-element', 'button']
will find the button, but a usual CSS selector 'open-shadow-root-element button'
will not.
{
"exists": ElementSelector
}
Returns true if the given selector matches one or more elements.
{
"visible": ElementSelector,
"check": "any" | "all" | "none"
}
Returns true if elements matched by ElementSelector are currently visible on the page. If check
is all
, every element must be visible. If check
is none
, no element should be visible. Visibility check is a CSS-based heuristic.
{
"waitFor": ElementSelector,
"timeout": 1000
}
Waits until selector
exists in the page. After timeout
ms the step fails.
{
"waitForVisible": ElementSelector,
"timeout": 1000,
"check": "any" | "all" | "none"
}
Waits until element is visible in the page. After timeout
ms the step fails.
{
"click": ElementSelector,
"all": true | false,
}
Click on an element returned by selector
. If all
is true
, all matching elements are clicked. If all
is false
, only the first returned value is clicked.
{
"waitForThenClick": ElementSelector,
"timeout": 1000,
"all": true | false
}
Combines waitFor
and click
.
{
"wait": 1000,
}
Wait for the specified number of milliseconds.
{
"hide": "CSS selector",
"method": "display" | "opacity"
}
Hide the elements matched by the selectors. method
defines how elements are hidden: "display" sets display: none
, "opacity" sets opacity: 0
. Method is "display" by default. Note that only a single string CSS selector is supported here, not an array.
{
"eval": "SNIPPET_ID"
}
Evaluates a code snippet in the context of the page. The rule is considered successful if it evaluates to a truthy value. Snippets have to be explicitly defined in snippets.ts. Eval rules are not 100% reliable because they can be affected by the page scripts, or blocked by a CSP policy on the page. Therefore, they should only be used as a last resort when none of the other rules are sufficient.
{
"if": { "exists": ElementSelector },
"then": [
{ "click": ".button1" },
{ "click": ".button3" }
],
"else": [
{ "click": ".button2" }
]
}
Allows to do conditional branching in JSON rules. The if
section can contain either a "visible" or "exists" rule. Depending on the result of that rule, then
or else
sequences will be executed. else
section is optional.
The "if" rule is considered successful as long as all rules inside the chosen branch are successful. The other branch, as well as the result of the condition itself, do not affect the result of the whole rule.
{
"any": [
{ "exists": ".button1" },
{ "exists": ".button2" }
]
}
Evaluates a list of steps in order. If any return true (success), then the step returns true. If all steps return false, the any
step returns false.
All rules can include the "optional": true
to ignore failure.
See this document for more details on internal APIs.
MPLv2.
To test the extension / addon with Firefox, open the about:debugging
, navigate to "This Firefox" on the menu and under "Temporary Extensions" click on "Load Temporary Addon". Select the manifest.json
file from the dist/firefox
directory. You will need to build the extension before as described above. The extension should then be active and you can test it manually by simply visiting websites.