The Accessibility Company

Tac.list Sites

List sites in the authenticated TAC account. When the token is locked to one site, only that site is returned. openIssueCount / toFixCount is confirmed defects only; needsReviewCount is the separate axe-incomplete queue. Do not treat needsReviewCount as a fix backlog.

Tac.list Open Issues

List issues for the account. Default stream is `fix` (confirmed defects only — axe violations and human-confirmed incompletes), ranked confirmed-work first. Do NOT pick the first row of stream=`all` or `review` and start writing code: those include axe "needs review" incompletes (gradient backgrounds, short text, pseudo-elements) that are usually noise. For the review queue use stream=`review` or tac.listNeedsReviewGroups. Filters: siteId, ruleId, minImpact, stream, limit, offset. Each row inclu

Tac.list Needs Review Groups

List the needs-review (axe incomplete) queue grouped by rule and incompleteReason (e.g. color-contrast × bgGradient). These are candidates, not confirmed defects — do not write code from this list. Use tac.dismissNeedsReviewGroup to clear a noise class, or confirm a group in the dashboard when a human has verified it is real. Filters: siteId.

Tac.dismiss Needs Review Group

Dismiss every issue in a needs-review group (ruleId + incompleteReason) as false-positive (or another ignore reason). Use after listNeedsReviewGroups when a class is noise — shortTextContent, unresolved bgGradient, pseudoContent — not for confirmed violations. Write scope required. Groups larger than 100 members may need a second call (truncated=true).

Tac.get Issue With Fix Context

Fetch one issue with everything an AI agent needs to write a fix: rule help URL, representative finding (selector + html snippet + failure summary), affected pages, a verify-fix hint, agent memory (priorDecisions, verificationHistory, fixHistory, regression signal), AND any AI context already generated in the dashboard (summary + before/after fix suggestion under aiContext). Pass findingsLimit > 0 to also get every failing occurrence's selector + snippet for template bugs with per-instance marku

Tac.get Decisions

Load the workspace's accumulated judgement before a session: the active ignore ledger (by-design / false-positive / wont-fix decisions) plus recently-reported fixes. Read this first so an agent doesn't re-flag or re-fix something already settled. Site scope is enforced server-side. Read scope.

Tac.get Compliance Checklist

Read a site's per-criterion WCAG compliance checklist: combined automated + human-review state for every criterion in the site's standard, with a progress rollup. Filter by status (failing / needs_review / compliant / not_applicable) to pull just the rows that need attention. Read scope.

Tac.mark Issue In Progress

Set an issue status to in_progress so the dashboard shows that an agent / human is actively working on it. Records the optional note on the activity timeline. Write scope required.

Tac.mark Issue Fixed

Set an issue status to fixed_pending_verification after pushing a fix. Use AFTER a passing tac.verifyFixAtUrl (or once the fix is deployed). The issue only becomes verified when a full re-scan confirms the rule no longer fires — an agent can never self-certify. Optional fixedBy records the PR / commit to the Agent Memory Layer. Write scope required.

Tac.mark Issue Ignored

Persist an issue-level ignore so the issue is hidden from the dashboard's default lens. Use for inspected false positives or by-design behaviour. Reason is required. Write scope required.

Tac.verify Fix At Url

Run a one-shot scan against an issue's affected URL (or one supplied by the agent) and return the identity-hash delta. Reports whether the issue's rule still fires (expectedFixedSummary). Rate-limited to 10/min per token. Write scope required.

Tac.list Review Checks

The site's manual review checks: one requirement on one page, with the plain question, what to try, what a pass looks like, the current answer and any proposed answer awaiting a person. Work through them with a browser and propose answers with tac.proposeReviewCheck. Filters: pageUrl, criterion, openOnly (default true). Read scope.

Tac.propose Review Check

Propose an answer on one manual review check (from tac.listReviewChecks) with a plain summary of what you did and saw, plus evidence: text items or https screenshot URLs. A person accepts or disagrees in the Review tab; nothing is recorded until they do. Write scope.