Skip to main content
Glama

Walk the page by keyboard

keyboard_walk
Read-only

Use this when the question is whether the site actually WORKS for a keyboard-only or switch-device user — tab order, focus traps, whether the focus ring is visible at all — which no stored scan and no static markup check can answer. Run it before claiming keyboard accessibility for a page, and again after deploying any tab-order, focus or skip-link fix. It tabs through the page the way a keyboard-only user does and reports what breaks: focus traps, elements with no visible focus indicator, focus hidden behind sticky headers, controls that cannot be reached at all, controls that ignore Enter and Space, and tab order that contradicts the visual or document order. Focus indicators are measured by comparing rendered pixels rather than CSS, because indicators come from outline, box-shadow, border and background alike. Findings carry a confidence: the order rules in particular are nudges for a human to confirm, not assertions. A focus-obscured finding names the element doing the covering, once per occluder, and counts distinct covered elements — with stops carrying the tab-stop total when the ring revisits them. A focus-trap finding that may be an open modal behaving correctly says so and says what to check by hand. The response states the viewport the walk was performed in, because the obscured and offscreen results are geometry, and endedBy says which of three things stopped the walk: the ring closing ("ring-closed"), the step ceiling ("max-steps") or the run's time budget ("time-budget"). Only the first means the ring was fully walked. Each finding's selectors array holds up to 5 examples, with count the true total and selectorsTruncated set when more were cut; full: true returns them all from the stored run. A run that lands on a bot-protection interstitial rather than the page is reported as void, never as clean, and a void run carries no stops or findings at all. Where the interstitial title is contradicted by the walk itself, the run stands and carries challengeSuspected instead: trust the walk, not the title. Read-only.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
urlNoPage to walk. Must belong to the website above. Defaults to its homepage when omitted.
fullNoReturn every selector each finding carries instead of the first 5. Read from the result this run already stored — it does NOT walk the page again and costs nothing against any allowance. This is the way to resolve selectorsOmitted: re-running the tool to see a truncated list spends a second browser check on work already done.
websiteYesThe website domain as registered in Inclusify, e.g. "example.com".
maxStepsNoCeiling on Tab presses (default 150). If the ring does not close within it, that is reported as unfinished rather than as a trap. The step ceiling is not the only limit: the whole run has a 45s budget on the worker, and the walk stops early if it runs out, reported as endedBy: "time-budget". Raising this is cheap in itself — a stop costs about 12ms — but the budget is mostly spent on the page load and the focus-indicator screenshots, so a run that ended on time will not walk further just because the ceiling is higher.
includeRingNoInclude the tab ring: one entry per stop, in order, with its name, role and whether focus was visible. Off by default — on a large page it is long, and the findings are usually what you want.

TDQS

A5/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description goes far beyond the read-only annotation: it discloses confidence levels, pixel-based focus measurement, endedBy semantics, void runs vs challengeSuspected, viewport dependence, selector truncation, and modal-trap caveats. There is no contradiction between the description and the annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is long but dense with non-redundant, operational details. It is front-loaded with the most important usage guidance and then systematically covers behavioral nuances, edge cases, and output semantics, so every sentence earns its place.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With no output schema present, the description fully compensates by explaining what findings look like, what endedBy values mean, how selectors and counts work, and how void and challengeSuspected runs are reported. It gives the agent everything needed to interpret results correctly.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Although the schema already covers 100% of parameters, the description adds meaningfully to them: it explains that full does not re-walk the page and costs nothing, maxSteps interacts with a 45s budget and endedBy, url defaults to the homepage, and includeRing returns one entry per stop. This is substantial value beyond the schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific purpose: testing whether a site actually works for keyboard-only or switch-device users, covering tab order, focus traps, and visible focus indicators. It also distinguishes itself from stored scans and static markup checks, making the tool's unique value immediately clear.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It gives explicit 'Use this when' guidance, tells the agent to run it before claiming keyboard accessibility and again after tab-order or skip-link fixes, and explains why static checks can't substitute for it. This is clear selection guidance with no ambiguity.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

A4.3/5.0
Disambiguation4/5

The descriptions do exceptional cross-referencing work, explicitly separating near-neighbor pairs (add_website vs add_domain, list_alt_findings vs check_page_alt_text, plan_options vs billing_link, widget_status vs widget_usage). A few clusters remain that an agent could confuse without reading carefully, notably site_overview vs compliance_status (both report statement existence and scan-record state) and crawl_summary vs list_monitored_pages vs site_overview (all touch coverage numbers). Overall, distinct purposes are clearly delineated despite the large surface.

Naming Consistency4/5

All names are lowercase snake_case with strong family patterns: list_* (5 tools), add_* (3), set_* (7), plus org_* and *_history pairs. The main inconsistency is the mix of verb-led names (list_violations, set_slack_channel, start_crawl) with noun-led read names (site_overview, compliance_status, widget_usage, next_steps), but the noun-led names follow a coherent 'what it returns' vocabulary (status, summary, history, overview, rollup). Minor deviations rather than chaos.

Tool Count3/5

36 tools is heavy and sits above the 25-tool threshold where agent navigation starts to degrade, but the server covers a genuinely broad domain: website lifecycle, monitoring, four finding types, four live-audit tools, seven config setters, org rollups, billing, and CI. Most tools earn their place and none are duplicates, but several could plausibly be merged (set_slack_channel/set_slack_events/set_monitoring_settings into one notifications tool; list_violations/list_alt_findings/list_content_findings with a filter). The count is on the edge of unwieldy for an agent's tool-selection step.

Completeness3/5

The read/audit/analysis side is rich and well-covered: findings, history, live checks, org rollups, coverage, and validation all have tools. However, the write side is one-directional: add_monitored_pages is explicitly add-only, and there is no remove_website, remove_domain, or remove_monitored_pages, so teardown and 'stop monitoring this page' requests hit dead ends that the descriptions acknowledge belong to the panel. Statement content writing and widget installation are also panel/browser-only by design, which is documented but still leaves those operations outside the agent's reach.