Community recipe · Tool workflows

Create a tagged PDF candidate for accessibility review

Source-backed example. Practical decision guidance is proposed and untested; this is not a full implementation guide.

The problem

An untagged or poorly structured PDF may be difficult to navigate, while automated tagging alone cannot establish meaningful accessibility.

Agent job

Preserve the original, classify and tag a working copy, inspect marked content, use approved OCR or fallback steps and stage a candidate.

Expected output

A tagged PDF candidate and status entry for human accessibility checking.

Where this framing comes from

Core problem: Editorial purpose framing
Agent job: Editorial purpose framing
Expected output: Editorial purpose framing
Relevant job roles: Editorial purpose framing
Benefit type: Editorial purpose framing

Editorial framing and outcome limits

Describes the task’s purpose and intended output; measured benefit, safe unsupervised operation and successful delivery are not established by this framing.

Editorial synthesis based on the linked source. This framing does not add evidence of successful delivery or results.

Case-specific decision guidance · untested

Case-specific decision guidance and first checks are original proposals. The linked source supports the attributed task; this added guidance has not been executed.

Read the attributed source evidence and original caveats below ↓

Why use this

A document owner may need a PDF that can be navigated meaningfully rather than simply displayed on screen. Automated tags can create the appearance of progress while leaving reading order, tables or image descriptions unusable.
This job could prepare a candidate with its unresolved accessibility questions visible and the original preserved. A staged review artifact helps a human evaluator focus on semantic problems that a tag count or score cannot settle.

Why an agent / simpler alternative

Useful judgment: Classify document structure, identify where automated tagging is insufficient and propose targeted human checks.
Simpler alternative: Recreate a simple document from its accessible source file when that source is available.
Unsuitable when: Cloud processing lacks governance approval, original preservation is impossible or full compliance is expected without human review.

Proposed inputs

Name
Authorized PDF copy
Owner
Document owner
Minimum scope
Complete file and permitted processing scope
Quality check
Check text layer, structure and sensitive content
Name
Review context
Owner
Accessibility owner
Minimum scope
Audience, module naming and intended use
Quality check
Identify tables, equations and meaningful images

Access and setup — proposed or unknown

Proposed minimum: Inspect a copy and plan a candidate/status entry; no tagging execution, OCR installation or cloud upload.
Compatibility unknowns: Current tagging/OCR tools, cloud processing terms, licensing and exact runtime compatibility are unknown.
External actions requiring approval: Cloud uploads require data-governance approval; installations, external sharing and learning-platform uploads need separate authorization.

Expected deliverable and review checks

Proposed artifact: Candidate preparation plan with original-preservation record, structural issues and human accessibility-check checklist.
Critical checks: Inspect meaningful reading order, table relationships, equations and descriptions; retain canonical limits and licensing caution.
Missing output evidence: This overlay adds no new output inspection, execution or validation; consult the referenced canonical source outcome limits.

Failure signals and proposed responses

Signal
Tags exist but reading order is wrong
Why it matters
A numerical indicator hides an unusable experience
Proposed response
Review the actual sequence and propose manual correction.
Signal
OCR changes a mathematical symbol
Why it matters
Readable text becomes semantically inaccurate
Proposed response
Compare the expression visually and flag specialist review.

Effort and cost — unknown

Setup work: Classify document complexity and permitted processing routes.
Ongoing review: Human reviewers inspect semantics and record unresolved issues.
Unknowns: OCR accuracy, remediation effort, commercial reuse rights and costs are unknown.

First check — proposed, unexecuted

Status: proposed_unexecuted
Synthetic fixture: Invent a two-page document outline with a table, an image and an equation whose extracted symbol is ambiguous; create a review plan only.
Acceptance checks: Require semantic checks; preserve the original; distinguish tagged candidate from validated accessible output.
Stop conditions: Stop before cloud transfer or when meaningful content cannot be reconstructed reliably.

Limits

Use the canonical source outcome limits unchanged; candidate preparation is not an accessibility certification.

Tasks: Documents & files Security & compliance

Useful for: Designer Educator Administrative assistant

Editorial relevance from the title and source-supported actions; not evidence of the operator’s occupation.

Claude Code / CodexIndependent testing: No

Editorial candidate. This is source-backed reporting or a documented workflow. BotCurrent has not independently tested this task. A proposed Canadian adaptation is not evidence of a Canadian deployment or compliance certification.

Original source

https://github.com/tawhetsell/pdf-remediation-tool/blob/main/.claude/commands/remediate.md ↗

Attributed to: tawhetsell

Community recipe

Original action type: tool_action
What the source supports: Classify the PDF and preserve its original; tag a working copy, inspect real marked content, use OCR or a manual fallback when necessary, add metadata, and stage a candidate with a status entry for human accessibility checking.
Source outcome limits: Tagged candidate only; tag counts or an accessibility score do not establish semantic accessibility. Human reading-order, table, equation and image-description checks remain necessary. Clarify repository licensing before commercial code reuse.
Evidence attribution caveat: This community recipe describes a procedure; it does not prove successful execution or independently reproduced results.
Source approval description: Source requires data-governance approval before sending files to Adobe cloud APIs, prompts for naming/module context, and leaves Canvas/Ally upload and final checking to a person.
Named tools: Claude Code/Codex; PDF tagging and OCR pipeline; optional Adobe PDF Services; Canvas/Ally human review
Model or version: Unknown
Publication date: Unknown
Source last checked: 2026-10-14
Demonstrated in Canada: Unknown
Canadian availability: Unknown
Proposed Canadian adaptation: Proposed: handle authorized bilingual course/business PDFs, verify data residency and student-data permissions before any cloud service, and review applicable Canadian accessibility requirements.
Recommended approval boundaries: Approve document processing and any cloud transfer separately; preserve originals; have a document-accessibility specialist review reading order, headings, tables, equations and meaningful image descriptions.