[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7ADAC9JZFEJK2PKCN28R0MJ":3,"skill-same:01M7ADAC9JZFEJK2PKCN28R0MJ":142},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":42,"commitSha":43,"skillDirectory":44,"forksCount":45,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":6,"ownerDescription":48,"updatedAt":49,"install":50,"resources":63},"01M7ADAC9JZFEJK2PKCN28R0MJ","textract-document-deployer",null,"Deploys Amazon Textract document-analysis pipelines with production defaults: synchronous APIs (DetectDocumentText, AnalyzeDocument) for single-page real-time extraction, asynchronous APIs (StartDocumentAnalysis, StartDocumentTextDetection, StartExpenseAnalysis) for multipage PDFs up to 3000 pages, S3 document source and output config, SNS notification for async job completion, KMS encryption for input and output, feature selection (Forms, Tables, Queries, Signatures, Expense, Identity), bounding-box geometry and confidence-score handling, Lambda integration for real-time extraction, document splitting for oversized PDFs, service quotas, Comprehend integration for downstream NLP on extracted text. Emits a. Triggers: analyze document, detect document text, extract form fields, extract tables, textract queries, textract expense analysis, textract signature detection, textract identity documents, start document analysis, textract lambda, textract s3 output, textract kms encryption, textract sns notification.","jhkchan\u002Faws-agent-skills","jhkchan","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1083902?v=4","01M7A2SVQYP9GQTPTDXP5JKYQ4","Jacky Chan","01M7A2SVS117YWCTZABB75HZR1",4,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: textract-document-deployer\ndescription: 'Deploys Amazon Textract document-analysis pipelines with production defaults: synchronous APIs (DetectDocumentText, AnalyzeDocument) for single-page real-time extraction, asynchronous APIs (StartDocumentAnalysis, StartDocumentTextDetection, StartExpenseAnalysis) for multipage PDFs up to 3000 pages, S3 document source and output config, SNS notification for async job completion, KMS encryption for input and output, feature selection (Forms, Tables, Queries, Signatures, Expense, Identity), bounding-box geometry and confidence-score handling, Lambda integration for real-time extraction, document splitting for oversized PDFs, service quotas, Comprehend integration for downstream NLP on extracted text. Emits a. Triggers: analyze document, detect document text, extract form fields, extract tables, textract queries, textract expense analysis, textract signature detection, textract identity documents, start document analysis, textract lambda, textract s3 output, textract kms encryption, textract sns notification.'\nlicense: Apache-2.0\ncompatibility: 'Agent runtime that reads SKILL.md (Claude Code, Cursor, Windsurf, Codex, Gemini). For live deployment: AWS CLI v2 with textract access and an IAM role granting s3:GetObject on the input bucket, s3:PutObject on the output bucket, textract:Start* \u002F Get* \u002F Notify* actions, kms:Decrypt \u002F kms:GenerateDataKey on the KMS key, and sns:Publish if SNS notification is used. Works with Lambda runtimes (boto3 textract client), Step Functions orchestrations, Terraform aws_textract_* data sources, and...'\nmetadata:\n  domain: aws-cloudops\n  complexity: high\n  requires_llm: 'true'\n  phase: '1'\n  supports_pipeline: 'true'\n  entry_point: 'false'\n  family: AI\u002FML\n  task_type: deploy\n  skill_class: capability\n  lifecycle_status: active\n  verdict_shape: READY_TO_DEPLOY | PREREQUISITES_MISSING\n  version: 0.1.0\n  author: Jacky Chan — AWS Community Builder\n  tags: aws, textract, document-analysis, ml, deploy, provisioning, forms, tables, queries, expense, lambda, kms, sns\n  dependencies: aws-orchestrator\n  keywords: aws, textract, document analysis, ocr, forms, tables, queries, signatures, expense, identity documents, ml, deploy, provisioning, synchronous, asynchronous, lambda, kms, sns, s3, comprehend\n  when_to_use: Invoke when the user wants to extract text, forms, tables, signatures, expense, or identity data from documents (PDF, TIFF, PNG, JPEG) using Amazon Textract. Covers both synchronous APIs for single-page real-time extraction (DetectDocumentText, AnalyzeDocument) and asynchronous APIs for multipage batch extraction (StartDocumentAnalysis, StartExpenseAnalysis, StartDocumentTextDetection) with S3 input\u002Foutput, SNS completion notification, KMS encryption, Lambda integration, and Comprehend downstream NLP. Do NOT invoke for Amazon Transcribe (audio), Amazon Rekognition (image\u002Fvideo ML), or pure OCR via Textract DetectDocumentText without downstream analysis.\n---\n\n# Textract Document Deployer\n\nAn AWS CloudOps agent skill that deploys Amazon Textract document-\nanalysis pipelines with correct defaults. The skill walks the operator\nthrough feature selection (Forms, Tables, Queries, Signatures, Expense,\nIdentity), synchronous vs asynchronous API choice, S3 input\u002Foutput\nconfiguration, KMS encryption, SNS completion notifications, Lambda\nintegration for real-time extraction, document splitting for oversized\nPDFs, service quotas, and Comprehend downstream NLP, captures the\ndocument-processing topology, explains why each default matters, and\nemits a READY_TO_DEPLOY checklist with copy-pasteable verification\ncommands.\n\n## Activation keywords\n\nanalyze document, detect document text, extract form fields, extract\ntables, textract queries, textract expense analysis, textract signature\ndetection, textract identity documents, start document analysis, textract\nlambda, textract s3 output, textract kms encryption, textract sns\nnotification.\n\n## STRICT output contract\n\nWhen this skill is invoked with a Textract-deployment request\n(extract text\u002Fforms\u002Ftables\u002Fexpense\u002Fidentity from documents, set up async\nmultipage extraction, configure SNS\u002FKMS\u002FLambda integration, or a partial\nconfiguration), the agent MUST respond with the READY_TO_DEPLOY checklist\ndefined in the \"Output format\" section using the literal all-caps labels\n`TEXTRACT_PIPELINE:`, `VERDICT:`, `CHECKLIST:`, and\n`VERIFICATION_COMMANDS:`. Do NOT preface the checklist with prose,\nheadings, or disclaimers — emit the block as the first lines of the\nresponse. This contract is what assertion-based evals and downstream\nprovisioning pipelines rely on; deviating from the literal labels breaks\nautomation silently.\n\nIf any prerequisite is missing, the verdict is `PREREQUISITES_MISSING`\nwith a specific gap citation in the checklist (marked `[✗]`), and\n`READY_TO_DEPLOY` MUST NOT also appear.\n\n## Quick navigation\n\n| Section | When to read |\n|---|---|\n| Prerequisites | Always — verify before provisioning |\n| Step 1 — Feature selection (Forms, Tables, Queries, ...) | Core feature matrix |\n| Step 2 — Synchronous vs asynchronous API choice | API selection |\n| Step 3 — S3 document source and output config | Storage |\n| Step 4 — KMS encryption | Encryption |\n| Step 5 — SNS notification for async completion | Async notification |\n| Step 6 — Lambda integration for real-time extraction | Real-time path |\n| Step 7 — Bounding boxes and confidence scores | Geometry |\n| Step 8 — Document splitting for large PDFs | Quota handling |\n| Step 9 — Comprehend integration for downstream NLP | NLP chaining |\n| Step 10 — Identity documents and signatures | Specialized features |\n| Step 11 — Recent features | Latest |\n| NEVER do these things | Review before signing off |\n| Output format | The literal checklist template |\n| references\u002Fasync-and-s3.md | Async API + S3 output detail |\n| references\u002Fsync-and-lambda.md | Sync API + Lambda real-time detail |\n\n## Mindset\n\n**One-line takeaway:** Textract synchronous APIs\n(DetectDocumentText, AnalyzeDocument) work on a single page (or up to\nthe per-call page limit) for real-time use; asynchronous APIs\n(StartDocumentAnalysis, StartExpenseAnalysis,\nStartDocumentTextDetection) work on multipage PDFs up to 3000 pages\nwith S3 input, SNS notification, and an output config that writes\nstructured JSON to your own S3 bucket. Feature selection at\n`AnalyzeDocument` time — Forms, Tables, Queries, Signatures — is what\nturns raw OCR into structured business data.\n\nThree misconceptions dominate Textract misdesign at provisioning time:\n\n- **\"DetectDocumentText is sufficient for forms\u002Ftables.\"** It is not.\n  `DetectDocumentText` returns only raw text lines and words (with\n  geometry and confidence). To get form key-value pairs, table cells,\n  signatures, or natural-language query answers, you MUST use\n  `AnalyzeDocument` with the appropriate `FeatureTypes` (FORMS, TABLES,\n  QUERIES, SIGNATURES). A baseline model often reaches for\n  `DetectDocumentText` when the operator needs structured form data.\n\n- **\"The synchronous API handles multipage PDFs.\"** It does NOT.\n  `AnalyzeDocument` and `DetectDocumentText` accept a single document\n  (page). For multipage PDFs (up to 3000 pages), you MUST use the\n  asynchronous `StartDocumentAnalysis` \u002F `StartDocumentTextDetection` \u002F\n  `StartExpenseAnalysis` APIs with an S3 document source. Using the\n  sync API on a multipage PDF returns an error or processes only the\n  first page.\n\n- **\"Textract output is just the API response.\"** For async jobs, the\n  output is configurable. The `OutputConfig` parameter writes the\n  structured JSON to a customer-owned S3 bucket, and the\n  `S3ObjectPrefix` keeps results organized per job. Without\n  `OutputConfig`, results are returned only in the `Get*` API response\n  (which can be tens of MB for multipage documents).\n\n## Configuration dependency graph (novel heuristic)\n\nTextract configurations are NOT independent. The feature types selected\ndetermine the API choice; the API choice determines sync vs async; the\nasync path requires S3 input and (recommended) S3 output, SNS, and KMS.\nUse this graph to sequence provisioning.\n\n| Configuration | Hard dependencies (API error without) | Silent failure \u002F immutability | Enables downstream |\n|---|---|---|---|\n| IAM role (or direct IAM) | s3:GetObject on input bucket | Textract returns InvalidParameterException if S3 access is missing | Textract to read documents |\n| Synchronous AnalyzeDocument | Inline bytes (base64) OR S3 location; feature types specified | Limited to single-page or low-page documents; multipage PDFs rejected | real-time extraction in Lambda |\n| Asynchronous StartDocumentAnalysis | S3 location for input; S3 bucket in same region | without OutputConfig, results only available via Get* (paginated, large) | multipage processing up to 3000 pages |\n| OutputConfig (S3 output) | S3 bucket in same region as Textract job; s3:PutObject grant | bucket region MUST match Textract region | structured JSON results in your bucket |\n| KMS encryption | KMS key policy grants Textract service kms:GenerateDataKey + kms:Decrypt | KMS key must be in same region | encrypts input (if encrypted in S3) and output JSON |\n| SNS notification | SNS topic in same region; Textract service allowed sns:Publish | without SNS, must poll GetDocumentAnalysis | push notification on job completion |\n| Lambda integration | Lambda in same region; IAM includes textract:Start* \u002F Get* | Lambda synchronous API for single-page only | real-time extraction endpoint |\n| Comprehend NLP | Textract output (text or forms) extracted and passed to Comprehend | Comprehend has its own quotas and regions | entity detection, PII, classification on extracted text |\n| Document splitting | Customer-side logic (Textract does NOT split) | Textract enforces 3000-page hard limit per async job | oversized PDF processing |\n| Queries feature | AnalyzeDocument with FeatureTypes=[QUERIES] + QueriesConfig | Queries requires the QUERIES feature type AND a QueriesConfig block | natural-language questions on forms |\n| Expense analysis | StartExpenseAnalysis (async) or AnalyzeExpense (sync) | Expense uses a SEPARATE API family; Forms\u002FTables feature types do not apply | invoices\u002Freceipts structured line items |\n| Identity documents | AnalyzeID (sync, dedicated API) | Identity uses a SEPARATE API; Lending uses a SEPARATE async API | driver's license, passport fields |\n\n**The sync-vs-async choice row is the one a baseline model misses.**\nA naive deployment calls `DetectDocumentText` for everything. The\ncorrect heuristic recognizes that multipage PDFs and most production\nworkloads need the async API with S3 input, OutputConfig, and SNS. The\nprocedure below forces an explicit API choice.\n\n**Cross-dependency gotchas:**\n- S3 input bucket and OutputConfig bucket MUST be in the same AWS\n  region as the Textract API call. Cross-region S3 fails silently or\n  with a confusing InvalidS3Object exception.\n- KMS-encrypted input documents require the Textract service role (or\n  caller) to have kms:Decrypt on the encrypting key; the output JSON\n  uses a separate KMS key grant (kms:GenerateDataKey).\n- SNS topic MUST be in the same region as the Textract job. The role\n  needs sns:Publish on the topic ARN, and Textract's service principal\n  must be allowed.\n- The Queries feature is selected via `FeatureTypes=[QUERIES]` AND\n  requires a `QueriesConfig` block with one or more `Text` queries. A\n  missing QueriesConfig returns no query results.\n- Expense analysis and identity documents use SEPARATE API families\n  (`AnalyzeExpense` \u002F `StartExpenseAnalysis` and `AnalyzeID`). They are\n  NOT invoked through `AnalyzeDocument`.\n\n## Expert heuristic: choosing sync vs async\n\nA baseline model says \"use Textract.\" The correct heuristic recognizes\nthat Textract has two distinct operational shapes.\n\n```text\nDocument profile:\n  ├── Single page (PNG\u002FJPEG\u002FTIFF\u002FPDF), real-time, \u003C 30s SLA\n  │     → Synchronous API\n  │       DetectDocumentText  (raw text + geometry)\n  │       AnalyzeDocument     (text + Forms \u002F Tables \u002F Queries \u002F Signatures)\n  │       AnalyzeExpense      (invoices\u002Freceipts, single doc)\n  │       AnalyzeID           (driver's license \u002F passport)\n  │\n  ├── Multipage PDF (2-3000 pages), batch, minutes\n  │     → Asynchronous API with S3 input + OutputConfig + SNS + KMS\n  │       StartDocumentTextDetection  (raw text, multipage)\n  │       StartDocumentAnalysis       (Forms \u002F Tables \u002F Queries \u002F Signatures, multipage)\n  │       StartExpenseAnalysis        (invoices\u002Freceipts, multipage batch)\n  │\n  └── Over 3000 pages\n        → Customer-side document splitting into \u003C3000-page chunks\n          then StartDocumentAnalysis per chunk; aggregate downstream\n```\n\n**Key implication:** the async API is the production default for any\nmultipage document. Lambda real-time extraction is only for single-page\nor low-page documents where the 15-minute (or configured) Lambda timeout\nis not exceeded.\n\n## Expert heuristic: feature-type selection\n\nFeature types are selected at `AnalyzeDocument` time. Each adds\nstructured output on top of raw OCR.\n\n```text\nUse case                                → FeatureTypes         → API\n─────────────────────────────────────────────────────────────────────────\nRaw text extraction (lines, words)     → (none — use Detect)  → DetectDocumentText\nForm key-value pairs (labels+values)   → [FORMS]              → AnalyzeDocument\nTabular data (rows, columns, headers)  → [TABLES]             → AnalyzeDocument\nNatural-language questions on forms    → [QUERIES]            → AnalyzeDocument\nSignature locations on page            → [SIGNATURES]         → AnalyzeDocument\nForms + Tables together                → [FORMS, TABLES]      → AnalyzeDocument\nForms + Tables + Queries               → [FORMS, TABLES, QUERIES]\nInvoice \u002F receipt line items           → (separate API)       → AnalyzeExpense \u002F StartExpenseAnalysis\nDriver's license \u002F passport            → (separate API)       → AnalyzeID\n```\n\n**The Queries feature lets you ask natural-language questions of\nforms.** Example: \"What is the policy number?\" returns the value and\nits bounding box, even if the form layout varies. This is more\nflexible than FORMS key-value extraction, which depends on matching\nlabels.\n\n## Expert heuristic: async API supports up to 3000 pages per job\n\nThe async APIs (`StartDocumentAnalysis`, `StartDocumentTextDetection`,\n`StartExpenseAnalysis`) support multipage PDFs up to 3000 pages per\njob (hard limit). The sync APIs support a single page (or a small\nnumber of pages depending on the API and document type). For oversized\nPDFs, split the document customer-side before invoking Textract.\n\n```text\nPDF page count:\n  ├── 1 page            → sync OK (Lambda real-time)\n  ├── 2-3000 pages      → async (StartDocumentAnalysis)\n  └── >3000 pages       → split into chunks of ≤3000 pages, async per chunk\n```\n\n**Key implication:** the 3000-page limit is a hard quota. Plan for\ndocument splitting in any document-management pipeline that handles\nlarge reports, contracts, or regulatory filings.\n\n## Expert heuristic: expense analysis is purpose-built for invoices and receipts\n\n`AnalyzeExpense` (sync) and `StartExpenseAnalysis` (async) use a\npurpose-built ML model for invoices and receipts. They return\nstructured fields (vendor, invoice number, total, line items) without\nrequiring Forms\u002FTables feature configuration.\n\n```text\nInvoice \u002F receipt processing decision:\n  ├── Need structured fields (vendor, total, line items)\n  │     → AnalyzeExpense \u002F StartExpenseAnalysis (purpose-built)\n  │       Returns: ExpenseDocuments[] with SummaryFields[] and LineItems[]\n  │\n  └── Need custom form layout extraction\n        → AnalyzeDocument with FORMS, TABLES, QUERIES\n```\n\n**Key implication:** do NOT use AnalyzeDocument with FORMS for invoices\nif you need vendor name, total, or line-item extraction. Use\nAnalyzeExpense — it returns these as structured fields with confidence\nscores.\n\n## Prerequisites (verify before provisioning)\n\nBefore emitting provisioning commands, verify these prerequisites. If\nany are missing, the verdict is **PREREQUISITES_MISSING**.\n\n| Prerequisite | Why it matters | How to verify |\n|---|---|---|\n| S3 input bucket exists (same region) | Textract must read documents from S3 in the same region | `aws s3api head-bucket --bucket \u003Cinput-bucket>` |\n| S3 output bucket exists (same region) | OutputConfig writes JSON to a customer-owned bucket | `aws s3api head-bucket --bucket \u003Coutput-bucket>` |\n| IAM grant: s3:GetObject on input | Textract service (or caller) must read documents | Review role policy |\n| IAM grant: s3:PutObject on output | Textract writes structured JSON to output bucket | Review role policy |\n| IAM grant: textract:Start*\u002FGet*\u002FNotify* | Required to invoke async APIs and retrieve results | Review IAM policy |\n| KMS key (if encryption enabled) | KMS-encrypted inputs require kms:Decrypt; outputs use kms:GenerateDataKey | `aws kms describe-key --key-id \u003Ckey-id>` |\n| KMS key policy grants Textract | Textract service principal needs kms:GenerateDataKey + kms:Decrypt | Review key policy |\n| SNS topic (if async + notification) | SNS topic in same region; Textract needs sns:Publish | `aws sns get-topic-attributes --topic-arn \u003Carn>` |\n| Document format supported | PDF, TIFF, PNG, JPEG | Confirm file type |\n| Document under 3000 pages (async) | Hard limit per async job | Verify page count, plan splitting |\n| Feature type selected | Determines structured output | Choose from FORMS\u002FTABLES\u002FQUERIES\u002FSIGNATURES |\n\nIf any prerequisite is missing, output `VERDICT: PREREQUISITES_MISSING`\nand cite the specific gap.\n\n## Step 1 — Feature selection (Forms, Tables, Queries, Signatures, Expense, Identity)\n\n`AnalyzeDocument` accepts a `FeatureTypes` list. Multiple features can\nbe combined (e.g., `[FORMS, TABLES, QUERIES]`).\n\n| Feature | What it adds | API |\n|---|---|---|\n| FORMS | Key-value pairs (field labels + values) with confidence | AnalyzeDocument |\n| TABLES | Tabular structure (rows, columns, headers, merged cells) | AnalyzeDocument |\n| QUERIES | Natural-language question answering on forms (with QueriesConfig) | AnalyzeDocument |\n| SIGNATURES | Signature locations (bounding boxes) on the page | AnalyzeDocument |\n| LAYOUT | Reading order and structural layout elements | AnalyzeDocument |\n| (Expense) | Invoice\u002Freceipt structured fields — separate API | AnalyzeExpense \u002F StartExpenseAnalysis |\n| (Identity) | Driver's license \u002F passport fields — separate API | AnalyzeID |\n\n**Choose feature types based on document type and downstream use:**\n- Invoices\u002Freceipts → Expense API (not AnalyzeDocument FORMS).\n- Custom forms → FORMS, optionally + QUERIES.\n- Spreadsheets → TABLES.\n- Contracts with signatures → FORMS + SIGNATURES.\n- Variable-layout forms → QUERIES (more flexible than FORMS).\n\n## Step 2 — Synchronous vs asynchronous API choice\n\n| Decision factor | Synchronous | Asynchronous |\n|---|---|---|\n| Document length | Single page (or low page count per API limits) | Up to 3000 pages |\n| Latency requirement | Real-time (\u003C 30s typical) | Minutes (batch) |\n| Input source | Inline bytes (base64) OR S3 Object | S3 Object (required) |\n| Output | API response only | OutputConfig (S3) recommended |\n| Notification | None (caller waits) | SNS topic (optional) |\n| Encryption | KMS via API parameters | KMS via API parameters |\n| APIs | DetectDocumentText, AnalyzeDocument, AnalyzeExpense, AnalyzeID | StartDocumentTextDetection, StartDocumentAnalysis, StartExpenseAnalysis |\n| Result retrieval | Direct in response | GetDocumentAnalysis \u002F GetExpenseAnalysis (paginated) OR OutputConfig S3 |\n| Lambda integration | Yes (single-page real-time) | Yes but with Step Functions for polling |\n\n**Rule of thumb:** if the document is more than 1-2 pages or the\nextraction takes more than a few seconds, use async.\n\n## Step 3 — S3 document source and output config\n\nOutputConfig + S3Prefix layout, same-region bucket requirement — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 4 — KMS encryption\n\nKMS key id, key-policy grants (kms:GenerateDataKey \u002F kms:Decrypt) — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 5 — SNS notification for async completion\n\nNotificationChannel RoleArn + SNSTopicArn, SNS message body fields — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 6 — Lambda integration for real-time extraction\n\nPython handler, minimum Lambda IAM policy, sync page-limit caveat — moved verbatim.\nFull detail: [Sync APIs and Lambda](references\u002Fsync-and-lambda.md).\n\n## Step 7 — Bounding boxes and confidence scores\n\nBoundingBox\u002FPolygon geometry, confidence-threshold filtering pattern — moved verbatim.\nFull detail: [Sync APIs and Lambda](references\u002Fsync-and-lambda.md).\n\n## Step 8 — Document splitting for large PDFs\n\n3000-page hard limit, chunk-and-submit pattern — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 9 — Comprehend integration for downstream NLP\n\nentities \u002F PII chaining, Comprehend size quotas — moved verbatim.\nFull detail: [Sync APIs and Lambda](references\u002Fsync-and-lambda.md).\n\n## Step 10 — Identity documents and signatures\n\nIdentity documents (driver's license, passport) use the dedicated\n`AnalyzeID` API (synchronous). It returns structured fields like\ndocument number, expiration, first\u002Flast name, address, date of birth.\n\n```bash\naws textract analyze-id \\\n  --document-pages '[{\"S3Object\":{\"Bucket\":\"my-bucket\",\"Name\":\"id\u002Fdrivers-license.jpg\"}}]' \\\n  --region us-east-1\n```\n\nSignatures are detected via `AnalyzeDocument` with\n`FeatureTypes=[\"SIGNATURES\"]`. Each signature is returned as a\nSIGNATURE Block with a bounding box.\n\n## Step 11 — Recent features\n\nQueries GA, Signatures GA, Expense CSV output, Lending, Layout, EU\u002FAPAC expansion — moved verbatim.\nFull detail: [Advanced patterns](references\u002Fadvanced-patterns.md).\n\n## NEVER do these things\n\n1. **NEVER use DetectDocumentText when you need forms, tables, queries,\n   signatures, or expense data.** DetectDocumentText returns only raw\n   text. Use AnalyzeDocument with the appropriate FeatureTypes, or\n   AnalyzeExpense \u002F AnalyzeID for specialized document types.\n\n2. **NEVER call the synchronous API on a multipage PDF expecting all\n   pages.** The sync APIs are designed for single-page (or low-page)\n   documents. Use StartDocumentAnalysis \u002F StartDocumentTextDetection\n   for multipage PDFs up to 3000 pages.\n\n3. **NEVER submit a PDF over 3000 pages without splitting.** The async\n   API hard-fails at 3001+ pages. Split customer-side into chunks of\n   ≤3000 pages and submit each as a separate job.\n\n4. **NEVER use cross-region S3 buckets for input or OutputConfig.**\n   Both the input bucket and the OutputConfig bucket MUST be in the\n   same region as the Textract API call. Cross-region S3 causes\n   InvalidS3Object or ProvisionedThroughputExceeded errors.\n\n5. **NEVER forget the notification role for SNS.** The\n   NotificationChannel requires a RoleArn that trusts\n   textract.amazonaws.com and has sns:Publish on the topic. Without\n   it, the API call fails.\n\n6. **NEVER use AnalyzeDocument FORMS for invoices and receipts.**\n   Invoices and receipts have a dedicated API (AnalyzeExpense \u002F\n   StartExpenseAnalysis) that returns structured vendor, total, and\n   line-item fields. AnalyzeDocument FORMS does not.\n\n7. **NEVER assume Textract output is small.** Multipage documents can\n   produce tens of MB of structured JSON. Use OutputConfig to write\n   results to S3 rather than fetching them all via Get* pagination.\n\n8. **NEVER skip the KMS key policy for Textract.** If the input\n   document is KMS-encrypted, Textract needs kms:Decrypt. If output\n   encryption is configured, Textract needs kms:GenerateDataKey. Both\n   must be granted to the Textract service principal.\n\n9. **NEVER assume confidence scores are binary (good\u002Fbad).** Set a\n   threshold appropriate to your use case (typically 50-90%). Route\n   low-confidence extractions to a human review queue.\n\n10. **NEVER confuse AnalyzeID with AnalyzeDocument.** Identity\n    documents (driver's license, passport) use the dedicated AnalyzeID\n    API. AnalyzeDocument does not return identity-document structured\n    fields.\n\n## Output format\n\n```text\nTEXTRACT_PIPELINE: \u003Cinput-s3-uri> → \u003Cfeature-types> → \u003Coutput-s3-uri>\nVERDICT: READY_TO_DEPLOY | PREREQUISITES_MISSING\nCHECKLIST:\n  [✓|✗] Input bucket: s3:\u002F\u002F\u003Cbucket>\u002F\u003Ckey> (region \u003Cregion>)\n  [✓|✗] Output bucket: s3:\u002F\u002F\u003Cbucket>\u002F\u003Cprefix> (region \u003Cregion>)\n  [✓|✗] API mode: Synchronous (AnalyzeDocument) | Asynchronous (StartDocumentAnalysis)\n  [✓|✗] Feature types: FORMS | TABLES | QUERIES | SIGNATURES | EXPENSE | IDENTITY\n  [✓|✗] Document format: \u003CPDF|TIFF|PNG|JPEG> (\u003Cpage-count> pages)\n  [✓|✗] Page-count check: PASS (\u003Cn> ≤ 3000) | FAIL (\u003Cn> > 3000 — split required)\n  [✓|✗] IAM — s3:GetObject on input: \u003Crole>\n  [✓|✗] IAM — s3:PutObject on output: \u003Crole>\n  [✓|✗] IAM — textract:Start*\u002FGet*: \u003Crole>\n  [✓|✗] KMS encryption: \u003Ckey-id> (input + output) | Disabled\n  [✓|✗] SNS notification: \u003Ctopic-arn> (RoleArn \u003Crole>) | Disabled\n  [✓|✗] Lambda integration: \u003Cfunction-name> (sync real-time) | Not used\n  [✓|✗] OutputConfig: s3:\u002F\u002F\u003Cbucket>\u002F\u003Cprefix>\u002F\u003Cjob-id>\u002F\n  [✓|✗] Queries config: \u003Cn> queries (if QUERIES feature) | Not used\n  [✓|✗] Confidence threshold: \u003Cthreshold>% (downstream review filter)\n  [✓|✗] Document splitting: not required | \u003Cn> chunks of ≤3000 pages\n  [✓|✗] Comprehend integration: \u003Cdetect-entities|detect-pii|classify> | Not used\nVERIFICATION_COMMANDS:\n  aws textract get-document-analysis --job-id \u003Cjob-id> --region \u003Cregion>\n  aws textract describe-document-text-detection --job-id \u003Cjob-id> --region \u003Cregion>  (if StartDocumentTextDetection)\n  aws s3 ls s3:\u002F\u002F\u003Coutput-bucket>\u002F\u003Cprefix>\u002F\u003Cjob-id>\u002F --recursive\n```\n\n### Worked example — async multipage Forms+Tables with SNS, KMS, OutputConfig\n\n```text\nTEXTRACT_PIPELINE: s3:\u002F\u002Fdoc-input\u002Finvoices\u002F2026-q3.pdf → FORMS,TABLES → s3:\u002F\u002Fdoc-output\u002Ftextract\u002F\nVERDICT: READY_TO_DEPLOY\nCHECKLIST:\n  [✓] Input bucket: s3:\u002F\u002Fdoc-input\u002Finvoices\u002F2026-q3.pdf (region us-east-1)\n  [✓] Output bucket: s3:\u002F\u002Fdoc-output\u002Ftextract\u002F (region us-east-1)\n  [✓] API mode: Asynchronous (StartDocumentAnalysis)\n  [✓] Feature types: FORMS, TABLES\n  [✓] Document format: PDF (1450 pages)\n  [✓] Page-count check: PASS (1450 ≤ 3000)\n  [✓] IAM — s3:GetObject on input: arn:aws:iam::123456789012:role\u002FTextractProcessingRole\n  [✓] IAM — s3:PutObject on output: arn:aws:iam::123456789012:role\u002FTextractProcessingRole\n  [✓] IAM — textract:StartDocumentAnalysis\u002FGetDocumentAnalysis: TextractProcessingRole\n  [✓] KMS encryption: arn:aws:kms:us-east-1:123456789012:key\u002Fabcd1234-... (input + output)\n  [✓] SNS notification: arn:aws:sns:us-east-1:123456789012:TextractComplete (RoleArn TextractNotificationRole)\n  [✓] OutputConfig: s3:\u002F\u002Fdoc-output\u002Ftextract\u002F\u003Cjob-id>\u002F\n  [✓] Confidence threshold: 75% (route \u003C75% to manual review)\n  [✓] Document splitting: not required\nVERIFICATION_COMMANDS:\n  aws textract get-document-analysis --job-id \u003Cjob-id> --region us-east-1\n  aws s3 ls s3:\u002F\u002Fdoc-output\u002Ftextract\u002F\u003Cjob-id>\u002F --recursive\n```\n\n## Error handling\n\nInvalidS3Object, KMS AccessDenied, IN_PROGRESS\u002FFAILED, ProvisionedThroughputExceeded, Lambda timeout, empty Queries — moved verbatim.\nFull detail: [Error handling](references\u002Ferror-handling.md).\n\n## References (load on demand)\n\n- [Async APIs and S3 output](references\u002Fasync-and-s3.md) — async API mechanics, OutputConfig layout, KMS key policy, SNS notification role, document splitting\n- [Sync APIs and Lambda](references\u002Fsync-and-lambda.md) — sync API limits, Lambda handler + IAM, bounding boxes and confidence filtering, Comprehend NLP chaining\n- [Advanced patterns](references\u002Fadvanced-patterns.md) — recent features (Queries, Signatures, Layout, Lending, OutputConfig CSV, region expansion)\n- [Error handling](references\u002Ferror-handling.md) — InvalidS3Object, KMS AccessDenied, job status, throughput, Lambda timeout, empty Queries\n\n## Domain\n\nAWS CloudOps \u002F Amazon Textract Document Analysis & OCR Pipeline\nProvisioning.\n\n## AWS documentation\n\n- **Textract Developer Guide** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002F\n- **AnalyzeDocument API** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_AnalyzeDocument.html\n- **StartDocumentAnalysis API** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_StartDocumentAnalysis.html\n- **DetectDocumentText API** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_DetectDocumentText.html\n- **OutputConfig** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_OutputConfig.html\n- **NotificationChannel** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_NotificationChannel.html\n- **Queries feature** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Fqueryresponse.html\n- **Expense Analysis** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Finvoices-receipts.html\n- **Identity Documents (AnalyzeID)** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Fidentitydocument.html\n- **Textract quotas** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Flimits-document.html\n- **KMS encryption** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Fsecurity_iam_encrypt-with-kms.html\n","# Textract Document Deployer\n\nAn AWS CloudOps agent skill that deploys Amazon Textract document-\nanalysis pipelines with correct defaults. The skill walks the operator\nthrough feature selection (Forms, Tables, Queries, Signatures, Expense,\nIdentity), synchronous vs asynchronous API choice, S3 input\u002Foutput\nconfiguration, KMS encryption, SNS completion notifications, Lambda\nintegration for real-time extraction, document splitting for oversized\nPDFs, service quotas, and Comprehend downstream NLP, captures the\ndocument-processing topology, explains why each default matters, and\nemits a READY_TO_DEPLOY checklist with copy-pasteable verification\ncommands.\n\n## Activation keywords\n\nanalyze document, detect document text, extract form fields, extract\ntables, textract queries, textract expense analysis, textract signature\ndetection, textract identity documents, start document analysis, textract\nlambda, textract s3 output, textract kms encryption, textract sns\nnotification.\n\n## STRICT output contract\n\nWhen this skill is invoked with a Textract-deployment request\n(extract text\u002Fforms\u002Ftables\u002Fexpense\u002Fidentity from documents, set up async\nmultipage extraction, configure SNS\u002FKMS\u002FLambda integration, or a partial\nconfiguration), the agent MUST respond with the READY_TO_DEPLOY checklist\ndefined in the \"Output format\" section using the literal all-caps labels\n`TEXTRACT_PIPELINE:`, `VERDICT:`, `CHECKLIST:`, and\n`VERIFICATION_COMMANDS:`. Do NOT preface the checklist with prose,\nheadings, or disclaimers — emit the block as the first lines of the\nresponse. This contract is what assertion-based evals and downstream\nprovisioning pipelines rely on; deviating from the literal labels breaks\nautomation silently.\n\nIf any prerequisite is missing, the verdict is `PREREQUISITES_MISSING`\nwith a specific gap citation in the checklist (marked `[✗]`), and\n`READY_TO_DEPLOY` MUST NOT also appear.\n\n## Quick navigation\n\n| Section | When to read |\n|---|---|\n| Prerequisites | Always — verify before provisioning |\n| Step 1 — Feature selection (Forms, Tables, Queries, ...) | Core feature matrix |\n| Step 2 — Synchronous vs asynchronous API choice | API selection |\n| Step 3 — S3 document source and output config | Storage |\n| Step 4 — KMS encryption | Encryption |\n| Step 5 — SNS notification for async completion | Async notification |\n| Step 6 — Lambda integration for real-time extraction | Real-time path |\n| Step 7 — Bounding boxes and confidence scores | Geometry |\n| Step 8 — Document splitting for large PDFs | Quota handling |\n| Step 9 — Comprehend integration for downstream NLP | NLP chaining |\n| Step 10 — Identity documents and signatures | Specialized features |\n| Step 11 — Recent features | Latest |\n| NEVER do these things | Review before signing off |\n| Output format | The literal checklist template |\n| references\u002Fasync-and-s3.md | Async API + S3 output detail |\n| references\u002Fsync-and-lambda.md | Sync API + Lambda real-time detail |\n\n## Mindset\n\n**One-line takeaway:** Textract synchronous APIs\n(DetectDocumentText, AnalyzeDocument) work on a single page (or up to\nthe per-call page limit) for real-time use; asynchronous APIs\n(StartDocumentAnalysis, StartExpenseAnalysis,\nStartDocumentTextDetection) work on multipage PDFs up to 3000 pages\nwith S3 input, SNS notification, and an output config that writes\nstructured JSON to your own S3 bucket. Feature selection at\n`AnalyzeDocument` time — Forms, Tables, Queries, Signatures — is what\nturns raw OCR into structured business data.\n\nThree misconceptions dominate Textract misdesign at provisioning time:\n\n- **\"DetectDocumentText is sufficient for forms\u002Ftables.\"** It is not.\n  `DetectDocumentText` returns only raw text lines and words (with\n  geometry and confidence). To get form key-value pairs, table cells,\n  signatures, or natural-language query answers, you MUST use\n  `AnalyzeDocument` with the appropriate `FeatureTypes` (FORMS, TABLES,\n  QUERIES, SIGNATURES). A baseline model often reaches for\n  `DetectDocumentText` when the operator needs structured form data.\n\n- **\"The synchronous API handles multipage PDFs.\"** It does NOT.\n  `AnalyzeDocument` and `DetectDocumentText` accept a single document\n  (page). For multipage PDFs (up to 3000 pages), you MUST use the\n  asynchronous `StartDocumentAnalysis` \u002F `StartDocumentTextDetection` \u002F\n  `StartExpenseAnalysis` APIs with an S3 document source. Using the\n  sync API on a multipage PDF returns an error or processes only the\n  first page.\n\n- **\"Textract output is just the API response.\"** For async jobs, the\n  output is configurable. The `OutputConfig` parameter writes the\n  structured JSON to a customer-owned S3 bucket, and the\n  `S3ObjectPrefix` keeps results organized per job. Without\n  `OutputConfig`, results are returned only in the `Get*` API response\n  (which can be tens of MB for multipage documents).\n\n## Configuration dependency graph (novel heuristic)\n\nTextract configurations are NOT independent. The feature types selected\ndetermine the API choice; the API choice determines sync vs async; the\nasync path requires S3 input and (recommended) S3 output, SNS, and KMS.\nUse this graph to sequence provisioning.\n\n| Configuration | Hard dependencies (API error without) | Silent failure \u002F immutability | Enables downstream |\n|---|---|---|---|\n| IAM role (or direct IAM) | s3:GetObject on input bucket | Textract returns InvalidParameterException if S3 access is missing | Textract to read documents |\n| Synchronous AnalyzeDocument | Inline bytes (base64) OR S3 location; feature types specified | Limited to single-page or low-page documents; multipage PDFs rejected | real-time extraction in Lambda |\n| Asynchronous StartDocumentAnalysis | S3 location for input; S3 bucket in same region | without OutputConfig, results only available via Get* (paginated, large) | multipage processing up to 3000 pages |\n| OutputConfig (S3 output) | S3 bucket in same region as Textract job; s3:PutObject grant | bucket region MUST match Textract region | structured JSON results in your bucket |\n| KMS encryption | KMS key policy grants Textract service kms:GenerateDataKey + kms:Decrypt | KMS key must be in same region | encrypts input (if encrypted in S3) and output JSON |\n| SNS notification | SNS topic in same region; Textract service allowed sns:Publish | without SNS, must poll GetDocumentAnalysis | push notification on job completion |\n| Lambda integration | Lambda in same region; IAM includes textract:Start* \u002F Get* | Lambda synchronous API for single-page only | real-time extraction endpoint |\n| Comprehend NLP | Textract output (text or forms) extracted and passed to Comprehend | Comprehend has its own quotas and regions | entity detection, PII, classification on extracted text |\n| Document splitting | Customer-side logic (Textract does NOT split) | Textract enforces 3000-page hard limit per async job | oversized PDF processing |\n| Queries feature | AnalyzeDocument with FeatureTypes=[QUERIES] + QueriesConfig | Queries requires the QUERIES feature type AND a QueriesConfig block | natural-language questions on forms |\n| Expense analysis | StartExpenseAnalysis (async) or AnalyzeExpense (sync) | Expense uses a SEPARATE API family; Forms\u002FTables feature types do not apply | invoices\u002Freceipts structured line items |\n| Identity documents | AnalyzeID (sync, dedicated API) | Identity uses a SEPARATE API; Lending uses a SEPARATE async API | driver's license, passport fields |\n\n**The sync-vs-async choice row is the one a baseline model misses.**\nA naive deployment calls `DetectDocumentText` for everything. The\ncorrect heuristic recognizes that multipage PDFs and most production\nworkloads need the async API with S3 input, OutputConfig, and SNS. The\nprocedure below forces an explicit API choice.\n\n**Cross-dependency gotchas:**\n- S3 input bucket and OutputConfig bucket MUST be in the same AWS\n  region as the Textract API call. Cross-region S3 fails silently or\n  with a confusing InvalidS3Object exception.\n- KMS-encrypted input documents require the Textract service role (or\n  caller) to have kms:Decrypt on the encrypting key; the output JSON\n  uses a separate KMS key grant (kms:GenerateDataKey).\n- SNS topic MUST be in the same region as the Textract job. The role\n  needs sns:Publish on the topic ARN, and Textract's service principal\n  must be allowed.\n- The Queries feature is selected via `FeatureTypes=[QUERIES]` AND\n  requires a `QueriesConfig` block with one or more `Text` queries. A\n  missing QueriesConfig returns no query results.\n- Expense analysis and identity documents use SEPARATE API families\n  (`AnalyzeExpense` \u002F `StartExpenseAnalysis` and `AnalyzeID`). They are\n  NOT invoked through `AnalyzeDocument`.\n\n## Expert heuristic: choosing sync vs async\n\nA baseline model says \"use Textract.\" The correct heuristic recognizes\nthat Textract has two distinct operational shapes.\n\n```text\nDocument profile:\n  ├── Single page (PNG\u002FJPEG\u002FTIFF\u002FPDF), real-time, \u003C 30s SLA\n  │     → Synchronous API\n  │       DetectDocumentText  (raw text + geometry)\n  │       AnalyzeDocument     (text + Forms \u002F Tables \u002F Queries \u002F Signatures)\n  │       AnalyzeExpense      (invoices\u002Freceipts, single doc)\n  │       AnalyzeID           (driver's license \u002F passport)\n  │\n  ├── Multipage PDF (2-3000 pages), batch, minutes\n  │     → Asynchronous API with S3 input + OutputConfig + SNS + KMS\n  │       StartDocumentTextDetection  (raw text, multipage)\n  │       StartDocumentAnalysis       (Forms \u002F Tables \u002F Queries \u002F Signatures, multipage)\n  │       StartExpenseAnalysis        (invoices\u002Freceipts, multipage batch)\n  │\n  └── Over 3000 pages\n        → Customer-side document splitting into \u003C3000-page chunks\n          then StartDocumentAnalysis per chunk; aggregate downstream\n```\n\n**Key implication:** the async API is the production default for any\nmultipage document. Lambda real-time extraction is only for single-page\nor low-page documents where the 15-minute (or configured) Lambda timeout\nis not exceeded.\n\n## Expert heuristic: feature-type selection\n\nFeature types are selected at `AnalyzeDocument` time. Each adds\nstructured output on top of raw OCR.\n\n```text\nUse case                                → FeatureTypes         → API\n─────────────────────────────────────────────────────────────────────────\nRaw text extraction (lines, words)     → (none — use Detect)  → DetectDocumentText\nForm key-value pairs (labels+values)   → [FORMS]              → AnalyzeDocument\nTabular data (rows, columns, headers)  → [TABLES]             → AnalyzeDocument\nNatural-language questions on forms    → [QUERIES]            → AnalyzeDocument\nSignature locations on page            → [SIGNATURES]         → AnalyzeDocument\nForms + Tables together                → [FORMS, TABLES]      → AnalyzeDocument\nForms + Tables + Queries               → [FORMS, TABLES, QUERIES]\nInvoice \u002F receipt line items           → (separate API)       → AnalyzeExpense \u002F StartExpenseAnalysis\nDriver's license \u002F passport            → (separate API)       → AnalyzeID\n```\n\n**The Queries feature lets you ask natural-language questions of\nforms.** Example: \"What is the policy number?\" returns the value and\nits bounding box, even if the form layout varies. This is more\nflexible than FORMS key-value extraction, which depends on matching\nlabels.\n\n## Expert heuristic: async API supports up to 3000 pages per job\n\nThe async APIs (`StartDocumentAnalysis`, `StartDocumentTextDetection`,\n`StartExpenseAnalysis`) support multipage PDFs up to 3000 pages per\njob (hard limit). The sync APIs support a single page (or a small\nnumber of pages depending on the API and document type). For oversized\nPDFs, split the document customer-side before invoking Textract.\n\n```text\nPDF page count:\n  ├── 1 page            → sync OK (Lambda real-time)\n  ├── 2-3000 pages      → async (StartDocumentAnalysis)\n  └── >3000 pages       → split into chunks of ≤3000 pages, async per chunk\n```\n\n**Key implication:** the 3000-page limit is a hard quota. Plan for\ndocument splitting in any document-management pipeline that handles\nlarge reports, contracts, or regulatory filings.\n\n## Expert heuristic: expense analysis is purpose-built for invoices and receipts\n\n`AnalyzeExpense` (sync) and `StartExpenseAnalysis` (async) use a\npurpose-built ML model for invoices and receipts. They return\nstructured fields (vendor, invoice number, total, line items) without\nrequiring Forms\u002FTables feature configuration.\n\n```text\nInvoice \u002F receipt processing decision:\n  ├── Need structured fields (vendor, total, line items)\n  │     → AnalyzeExpense \u002F StartExpenseAnalysis (purpose-built)\n  │       Returns: ExpenseDocuments[] with SummaryFields[] and LineItems[]\n  │\n  └── Need custom form layout extraction\n        → AnalyzeDocument with FORMS, TABLES, QUERIES\n```\n\n**Key implication:** do NOT use AnalyzeDocument with FORMS for invoices\nif you need vendor name, total, or line-item extraction. Use\nAnalyzeExpense — it returns these as structured fields with confidence\nscores.\n\n## Prerequisites (verify before provisioning)\n\nBefore emitting provisioning commands, verify these prerequisites. If\nany are missing, the verdict is **PREREQUISITES_MISSING**.\n\n| Prerequisite | Why it matters | How to verify |\n|---|---|---|\n| S3 input bucket exists (same region) | Textract must read documents from S3 in the same region | `aws s3api head-bucket --bucket \u003Cinput-bucket>` |\n| S3 output bucket exists (same region) | OutputConfig writes JSON to a customer-owned bucket | `aws s3api head-bucket --bucket \u003Coutput-bucket>` |\n| IAM grant: s3:GetObject on input | Textract service (or caller) must read documents | Review role policy |\n| IAM grant: s3:PutObject on output | Textract writes structured JSON to output bucket | Review role policy |\n| IAM grant: textract:Start*\u002FGet*\u002FNotify* | Required to invoke async APIs and retrieve results | Review IAM policy |\n| KMS key (if encryption enabled) | KMS-encrypted inputs require kms:Decrypt; outputs use kms:GenerateDataKey | `aws kms describe-key --key-id \u003Ckey-id>` |\n| KMS key policy grants Textract | Textract service principal needs kms:GenerateDataKey + kms:Decrypt | Review key policy |\n| SNS topic (if async + notification) | SNS topic in same region; Textract needs sns:Publish | `aws sns get-topic-attributes --topic-arn \u003Carn>` |\n| Document format supported | PDF, TIFF, PNG, JPEG | Confirm file type |\n| Document under 3000 pages (async) | Hard limit per async job | Verify page count, plan splitting |\n| Feature type selected | Determines structured output | Choose from FORMS\u002FTABLES\u002FQUERIES\u002FSIGNATURES |\n\nIf any prerequisite is missing, output `VERDICT: PREREQUISITES_MISSING`\nand cite the specific gap.\n\n## Step 1 — Feature selection (Forms, Tables, Queries, Signatures, Expense, Identity)\n\n`AnalyzeDocument` accepts a `FeatureTypes` list. Multiple features can\nbe combined (e.g., `[FORMS, TABLES, QUERIES]`).\n\n| Feature | What it adds | API |\n|---|---|---|\n| FORMS | Key-value pairs (field labels + values) with confidence | AnalyzeDocument |\n| TABLES | Tabular structure (rows, columns, headers, merged cells) | AnalyzeDocument |\n| QUERIES | Natural-language question answering on forms (with QueriesConfig) | AnalyzeDocument |\n| SIGNATURES | Signature locations (bounding boxes) on the page | AnalyzeDocument |\n| LAYOUT | Reading order and structural layout elements | AnalyzeDocument |\n| (Expense) | Invoice\u002Freceipt structured fields — separate API | AnalyzeExpense \u002F StartExpenseAnalysis |\n| (Identity) | Driver's license \u002F passport fields — separate API | AnalyzeID |\n\n**Choose feature types based on document type and downstream use:**\n- Invoices\u002Freceipts → Expense API (not AnalyzeDocument FORMS).\n- Custom forms → FORMS, optionally + QUERIES.\n- Spreadsheets → TABLES.\n- Contracts with signatures → FORMS + SIGNATURES.\n- Variable-layout forms → QUERIES (more flexible than FORMS).\n\n## Step 2 — Synchronous vs asynchronous API choice\n\n| Decision factor | Synchronous | Asynchronous |\n|---|---|---|\n| Document length | Single page (or low page count per API limits) | Up to 3000 pages |\n| Latency requirement | Real-time (\u003C 30s typical) | Minutes (batch) |\n| Input source | Inline bytes (base64) OR S3 Object | S3 Object (required) |\n| Output | API response only | OutputConfig (S3) recommended |\n| Notification | None (caller waits) | SNS topic (optional) |\n| Encryption | KMS via API parameters | KMS via API parameters |\n| APIs | DetectDocumentText, AnalyzeDocument, AnalyzeExpense, AnalyzeID | StartDocumentTextDetection, StartDocumentAnalysis, StartExpenseAnalysis |\n| Result retrieval | Direct in response | GetDocumentAnalysis \u002F GetExpenseAnalysis (paginated) OR OutputConfig S3 |\n| Lambda integration | Yes (single-page real-time) | Yes but with Step Functions for polling |\n\n**Rule of thumb:** if the document is more than 1-2 pages or the\nextraction takes more than a few seconds, use async.\n\n## Step 3 — S3 document source and output config\n\nOutputConfig + S3Prefix layout, same-region bucket requirement — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 4 — KMS encryption\n\nKMS key id, key-policy grants (kms:GenerateDataKey \u002F kms:Decrypt) — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 5 — SNS notification for async completion\n\nNotificationChannel RoleArn + SNSTopicArn, SNS message body fields — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 6 — Lambda integration for real-time extraction\n\nPython handler, minimum Lambda IAM policy, sync page-limit caveat — moved verbatim.\nFull detail: [Sync APIs and Lambda](references\u002Fsync-and-lambda.md).\n\n## Step 7 — Bounding boxes and confidence scores\n\nBoundingBox\u002FPolygon geometry, confidence-threshold filtering pattern — moved verbatim.\nFull detail: [Sync APIs and Lambda](references\u002Fsync-and-lambda.md).\n\n## Step 8 — Document splitting for large PDFs\n\n3000-page hard limit, chunk-and-submit pattern — moved verbatim.\nFull detail: [Async APIs and S3 output](references\u002Fasync-and-s3.md).\n\n## Step 9 — Comprehend integration for downstream NLP\n\nentities \u002F PII chaining, Comprehend size quotas — moved verbatim.\nFull detail: [Sync APIs and Lambda](references\u002Fsync-and-lambda.md).\n\n## Step 10 — Identity documents and signatures\n\nIdentity documents (driver's license, passport) use the dedicated\n`AnalyzeID` API (synchronous). It returns structured fields like\ndocument number, expiration, first\u002Flast name, address, date of birth.\n\n```bash\naws textract analyze-id \\\n  --document-pages '[{\"S3Object\":{\"Bucket\":\"my-bucket\",\"Name\":\"id\u002Fdrivers-license.jpg\"}}]' \\\n  --region us-east-1\n```\n\nSignatures are detected via `AnalyzeDocument` with\n`FeatureTypes=[\"SIGNATURES\"]`. Each signature is returned as a\nSIGNATURE Block with a bounding box.\n\n## Step 11 — Recent features\n\nQueries GA, Signatures GA, Expense CSV output, Lending, Layout, EU\u002FAPAC expansion — moved verbatim.\nFull detail: [Advanced patterns](references\u002Fadvanced-patterns.md).\n\n## NEVER do these things\n\n1. **NEVER use DetectDocumentText when you need forms, tables, queries,\n   signatures, or expense data.** DetectDocumentText returns only raw\n   text. Use AnalyzeDocument with the appropriate FeatureTypes, or\n   AnalyzeExpense \u002F AnalyzeID for specialized document types.\n\n2. **NEVER call the synchronous API on a multipage PDF expecting all\n   pages.** The sync APIs are designed for single-page (or low-page)\n   documents. Use StartDocumentAnalysis \u002F StartDocumentTextDetection\n   for multipage PDFs up to 3000 pages.\n\n3. **NEVER submit a PDF over 3000 pages without splitting.** The async\n   API hard-fails at 3001+ pages. Split customer-side into chunks of\n   ≤3000 pages and submit each as a separate job.\n\n4. **NEVER use cross-region S3 buckets for input or OutputConfig.**\n   Both the input bucket and the OutputConfig bucket MUST be in the\n   same region as the Textract API call. Cross-region S3 causes\n   InvalidS3Object or ProvisionedThroughputExceeded errors.\n\n5. **NEVER forget the notification role for SNS.** The\n   NotificationChannel requires a RoleArn that trusts\n   textract.amazonaws.com and has sns:Publish on the topic. Without\n   it, the API call fails.\n\n6. **NEVER use AnalyzeDocument FORMS for invoices and receipts.**\n   Invoices and receipts have a dedicated API (AnalyzeExpense \u002F\n   StartExpenseAnalysis) that returns structured vendor, total, and\n   line-item fields. AnalyzeDocument FORMS does not.\n\n7. **NEVER assume Textract output is small.** Multipage documents can\n   produce tens of MB of structured JSON. Use OutputConfig to write\n   results to S3 rather than fetching them all via Get* pagination.\n\n8. **NEVER skip the KMS key policy for Textract.** If the input\n   document is KMS-encrypted, Textract needs kms:Decrypt. If output\n   encryption is configured, Textract needs kms:GenerateDataKey. Both\n   must be granted to the Textract service principal.\n\n9. **NEVER assume confidence scores are binary (good\u002Fbad).** Set a\n   threshold appropriate to your use case (typically 50-90%). Route\n   low-confidence extractions to a human review queue.\n\n10. **NEVER confuse AnalyzeID with AnalyzeDocument.** Identity\n    documents (driver's license, passport) use the dedicated AnalyzeID\n    API. AnalyzeDocument does not return identity-document structured\n    fields.\n\n## Output format\n\n```text\nTEXTRACT_PIPELINE: \u003Cinput-s3-uri> → \u003Cfeature-types> → \u003Coutput-s3-uri>\nVERDICT: READY_TO_DEPLOY | PREREQUISITES_MISSING\nCHECKLIST:\n  [✓|✗] Input bucket: s3:\u002F\u002F\u003Cbucket>\u002F\u003Ckey> (region \u003Cregion>)\n  [✓|✗] Output bucket: s3:\u002F\u002F\u003Cbucket>\u002F\u003Cprefix> (region \u003Cregion>)\n  [✓|✗] API mode: Synchronous (AnalyzeDocument) | Asynchronous (StartDocumentAnalysis)\n  [✓|✗] Feature types: FORMS | TABLES | QUERIES | SIGNATURES | EXPENSE | IDENTITY\n  [✓|✗] Document format: \u003CPDF|TIFF|PNG|JPEG> (\u003Cpage-count> pages)\n  [✓|✗] Page-count check: PASS (\u003Cn> ≤ 3000) | FAIL (\u003Cn> > 3000 — split required)\n  [✓|✗] IAM — s3:GetObject on input: \u003Crole>\n  [✓|✗] IAM — s3:PutObject on output: \u003Crole>\n  [✓|✗] IAM — textract:Start*\u002FGet*: \u003Crole>\n  [✓|✗] KMS encryption: \u003Ckey-id> (input + output) | Disabled\n  [✓|✗] SNS notification: \u003Ctopic-arn> (RoleArn \u003Crole>) | Disabled\n  [✓|✗] Lambda integration: \u003Cfunction-name> (sync real-time) | Not used\n  [✓|✗] OutputConfig: s3:\u002F\u002F\u003Cbucket>\u002F\u003Cprefix>\u002F\u003Cjob-id>\u002F\n  [✓|✗] Queries config: \u003Cn> queries (if QUERIES feature) | Not used\n  [✓|✗] Confidence threshold: \u003Cthreshold>% (downstream review filter)\n  [✓|✗] Document splitting: not required | \u003Cn> chunks of ≤3000 pages\n  [✓|✗] Comprehend integration: \u003Cdetect-entities|detect-pii|classify> | Not used\nVERIFICATION_COMMANDS:\n  aws textract get-document-analysis --job-id \u003Cjob-id> --region \u003Cregion>\n  aws textract describe-document-text-detection --job-id \u003Cjob-id> --region \u003Cregion>  (if StartDocumentTextDetection)\n  aws s3 ls s3:\u002F\u002F\u003Coutput-bucket>\u002F\u003Cprefix>\u002F\u003Cjob-id>\u002F --recursive\n```\n\n### Worked example — async multipage Forms+Tables with SNS, KMS, OutputConfig\n\n```text\nTEXTRACT_PIPELINE: s3:\u002F\u002Fdoc-input\u002Finvoices\u002F2026-q3.pdf → FORMS,TABLES → s3:\u002F\u002Fdoc-output\u002Ftextract\u002F\nVERDICT: READY_TO_DEPLOY\nCHECKLIST:\n  [✓] Input bucket: s3:\u002F\u002Fdoc-input\u002Finvoices\u002F2026-q3.pdf (region us-east-1)\n  [✓] Output bucket: s3:\u002F\u002Fdoc-output\u002Ftextract\u002F (region us-east-1)\n  [✓] API mode: Asynchronous (StartDocumentAnalysis)\n  [✓] Feature types: FORMS, TABLES\n  [✓] Document format: PDF (1450 pages)\n  [✓] Page-count check: PASS (1450 ≤ 3000)\n  [✓] IAM — s3:GetObject on input: arn:aws:iam::123456789012:role\u002FTextractProcessingRole\n  [✓] IAM — s3:PutObject on output: arn:aws:iam::123456789012:role\u002FTextractProcessingRole\n  [✓] IAM — textract:StartDocumentAnalysis\u002FGetDocumentAnalysis: TextractProcessingRole\n  [✓] KMS encryption: arn:aws:kms:us-east-1:123456789012:key\u002Fabcd1234-... (input + output)\n  [✓] SNS notification: arn:aws:sns:us-east-1:123456789012:TextractComplete (RoleArn TextractNotificationRole)\n  [✓] OutputConfig: s3:\u002F\u002Fdoc-output\u002Ftextract\u002F\u003Cjob-id>\u002F\n  [✓] Confidence threshold: 75% (route \u003C75% to manual review)\n  [✓] Document splitting: not required\nVERIFICATION_COMMANDS:\n  aws textract get-document-analysis --job-id \u003Cjob-id> --region us-east-1\n  aws s3 ls s3:\u002F\u002Fdoc-output\u002Ftextract\u002F\u003Cjob-id>\u002F --recursive\n```\n\n## Error handling\n\nInvalidS3Object, KMS AccessDenied, IN_PROGRESS\u002FFAILED, ProvisionedThroughputExceeded, Lambda timeout, empty Queries — moved verbatim.\nFull detail: [Error handling](references\u002Ferror-handling.md).\n\n## References (load on demand)\n\n- [Async APIs and S3 output](references\u002Fasync-and-s3.md) — async API mechanics, OutputConfig layout, KMS key policy, SNS notification role, document splitting\n- [Sync APIs and Lambda](references\u002Fsync-and-lambda.md) — sync API limits, Lambda handler + IAM, bounding boxes and confidence filtering, Comprehend NLP chaining\n- [Advanced patterns](references\u002Fadvanced-patterns.md) — recent features (Queries, Signatures, Layout, Lending, OutputConfig CSV, region expansion)\n- [Error handling](references\u002Ferror-handling.md) — InvalidS3Object, KMS AccessDenied, job status, throughput, Lambda timeout, empty Queries\n\n## Domain\n\nAWS CloudOps \u002F Amazon Textract Document Analysis & OCR Pipeline\nProvisioning.\n\n## AWS documentation\n\n- **Textract Developer Guide** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002F\n- **AnalyzeDocument API** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_AnalyzeDocument.html\n- **StartDocumentAnalysis API** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_StartDocumentAnalysis.html\n- **DetectDocumentText API** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_DetectDocumentText.html\n- **OutputConfig** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_OutputConfig.html\n- **NotificationChannel** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002FAPI_NotificationChannel.html\n- **Queries feature** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Fqueryresponse.html\n- **Expense Analysis** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Finvoices-receipts.html\n- **Identity Documents (AnalyzeID)** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Fidentitydocument.html\n- **Textract quotas** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Flimits-document.html\n- **KMS encryption** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ftextract\u002Flatest\u002Fdg\u002Fsecurity_iam_encrypt-with-kms.html\n",{"name":5,"license":15,"metadata":24,"description":7,"compatibility":41},{"tags":25,"phase":26,"author":27,"domain":28,"family":29,"version":30,"keywords":31,"task_type":32,"complexity":33,"entry_point":34,"skill_class":35,"when_to_use":36,"dependencies":37,"requires_llm":38,"verdict_shape":39,"lifecycle_status":40,"supports_pipeline":38},"aws, textract, document-analysis, ml, deploy, provisioning, forms, tables, queries, expense, lambda, kms, sns","1","Jacky Chan — AWS Community Builder","aws-cloudops","AI\u002FML","0.1.0","aws, textract, document analysis, ocr, forms, tables, queries, signatures, expense, identity documents, ml, deploy, provisioning, synchronous, asynchronous, lambda, kms, sns, s3, comprehend","deploy","high","false","capability","Invoke when the user wants to extract text, forms, tables, signatures, expense, or identity data from documents (PDF, TIFF, PNG, JPEG) using Amazon Textract. Covers both synchronous APIs for single-page real-time extraction (DetectDocumentText, AnalyzeDocument) and asynchronous APIs for multipage batch extraction (StartDocumentAnalysis, StartExpenseAnalysis, StartDocumentTextDetection) with S3 input\u002Foutput, SNS completion notification, KMS encryption, Lambda integration, and Comprehend downstream NLP. Do NOT invoke for Amazon Transcribe (audio), Amazon Rekognition (image\u002Fvideo ML), or pure OCR via Textract DetectDocumentText without downstream analysis.","aws-orchestrator","true","READY_TO_DEPLOY | PREREQUISITES_MISSING","active","Agent runtime that reads SKILL.md (Claude Code, Cursor, Windsurf, Codex, Gemini). For live deployment: AWS CLI v2 with textract access and an IAM role granting s3:GetObject on the input bucket, s3:PutObject on the output bucket, textract:Start* \u002F Get* \u002F Notify* actions, kms:Decrypt \u002F kms:GenerateDataKey on the KMS key, and sns:Publish if SNS notification is used. Works with Lambda runtimes (boto3 textract client), Step Functions orchestrations, Terraform aws_textract_* data sources, and...","main","eeace51bd5dd1db2030729d83eeab988861553aa","skills\u002Ftextract-document-deployer",0,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills","https:\u002F\u002Fgithub.com\u002Fjhkchan","CTO & Co-Founder @Beever-AI  & @votee. CAISP. AWS Community Builder. Organizer@GDGCloudHK. MongoDB Champion. Technopreneur + Infornographer.","2026-09-24 08:54:10",{"source":8,"skillName":5,"commands":51,"prompt":61,"zipAvailable":62},[52,55,58],{"tool":53,"command":54},"npx","npx skills add jhkchan\u002Faws-agent-skills --skill textract-document-deployer",{"tool":56,"command":57},"bunx","bunx skills add jhkchan\u002Faws-agent-skills --skill textract-document-deployer",{"tool":59,"command":60},"pnpm","pnpm dlx skills add jhkchan\u002Faws-agent-skills --skill textract-document-deployer","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002FSKILL.md",true,[64,70,74,78,82,86,90,94,98,102,106,110,114,118,122,126,130,134,138],{"path":65,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":68,"sourceHtmlUrl":69},"eval\u002Ftest-cases.yaml","REFERENCE","参考文档",6142,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Feval\u002Ftest-cases.yaml",{"path":71,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":72,"sourceHtmlUrl":73},"evals\u002Fbaselines\u002Fasync-forms-tables-sns-kms.md",676,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fbaselines\u002Fasync-forms-tables-sns-kms.md",{"path":75,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":76,"sourceHtmlUrl":77},"evals\u002Fbaselines\u002Fexpense-analysis-invoices.md",713,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fbaselines\u002Fexpense-analysis-invoices.md",{"path":79,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":80,"sourceHtmlUrl":81},"evals\u002Fbaselines\u002Flambda-sync-realtime.md",959,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fbaselines\u002Flambda-sync-realtime.md",{"path":83,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":84,"sourceHtmlUrl":85},"evals\u002Fbaselines\u002Fmissing-s3-getobject-permission.md",792,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fbaselines\u002Fmissing-s3-getobject-permission.md",{"path":87,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":88,"sourceHtmlUrl":89},"evals\u002Fbaselines\u002Fqueries-natural-language.md",719,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fbaselines\u002Fqueries-natural-language.md",{"path":91,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":92,"sourceHtmlUrl":93},"evals\u002Feval_config.yaml",1803,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Feval_config.yaml",{"path":95,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":96,"sourceHtmlUrl":97},"evals\u002Fevals.json",10002,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fevals.json",{"path":99,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":100,"sourceHtmlUrl":101},"evals\u002Fprompts\u002Fasync-forms-tables-sns-kms.md",700,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fprompts\u002Fasync-forms-tables-sns-kms.md",{"path":103,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":104,"sourceHtmlUrl":105},"evals\u002Fprompts\u002Fexpense-analysis-invoices.md",579,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fprompts\u002Fexpense-analysis-invoices.md",{"path":107,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":108,"sourceHtmlUrl":109},"evals\u002Fprompts\u002Flambda-sync-realtime.md",509,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fprompts\u002Flambda-sync-realtime.md",{"path":111,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":112,"sourceHtmlUrl":113},"evals\u002Fprompts\u002Fmissing-s3-getobject-permission.md",564,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fprompts\u002Fmissing-s3-getobject-permission.md",{"path":115,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":116,"sourceHtmlUrl":117},"evals\u002Fprompts\u002Fqueries-natural-language.md",592,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fevals\u002Fprompts\u002Fqueries-natural-language.md",{"path":119,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":120,"sourceHtmlUrl":121},"examples\u002FREADME.md",6274,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fexamples\u002FREADME.md",{"path":123,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":124,"sourceHtmlUrl":125},"references\u002Fadvanced-patterns.md",1632,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Freferences\u002Fadvanced-patterns.md",{"path":127,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":128,"sourceHtmlUrl":129},"references\u002Fasync-and-s3.md",10901,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Freferences\u002Fasync-and-s3.md",{"path":131,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":132,"sourceHtmlUrl":133},"references\u002Ferror-handling.md",1560,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Freferences\u002Ferror-handling.md",{"path":135,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":136,"sourceHtmlUrl":137},"references\u002Fsync-and-lambda.md",12693,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Freferences\u002Fsync-and-lambda.md",{"path":139,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":140,"sourceHtmlUrl":141},"skill.usf.yaml",4516,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftextract-document-deployer\u002Fskill.usf.yaml",[143,150,157,164,171,178],{"skillNo":144,"name":145,"nameZh":6,"description":146,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":147,"categories":148,"keywords":149,"publishedAt":19},"01M7ADVD60X4PCAPKSE58P9NDC","xray-tracing-deployer","Deploys AWS X-Ray distributed tracing with production-grade configuration: X-Ray daemon deployment (EC2 systemd, ECS sidecar, EKS DaemonSet, Lambda built-in), SDK instrumentation (Java, Python, Node.js, Go, .NET) with annotations and metadata, sampling rules (default reservoir + rate vs custom rules with URL \u002F method \u002F service-name predicates), service map generation, trace query and retrieval, groups and insights, CloudWatch ServiceLens integration, IAM permissions (PutTraceSegments, GetSamplingRules), and latest features (Lambda Powertools tracing, OpenTelemetry \u002F ADOT Collector support, W3C trace context). Emits a READY_TO_DEPLOY checklist with every configuration item verified. Use when enabling X-Ray on a new application, adding tracing to ECS \u002F EKS \u002F Lambda, validating a sampling configuration, or generating daemon deployment CLI and IaC templates. Triggers: X-Ray, distributed tracing, X-Ray daemon, sampling rules, service map, X-Ray SDK, OpenTelemetry, ADOT, Lambda Powertools, CloudWatch ServiceLens.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fxray-tracing-deployer\u002FSKILL.md",[],[],{"skillNo":151,"name":152,"nameZh":6,"description":153,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":154,"categories":155,"keywords":156,"publishedAt":19},"01M7ADTVJ7DCBM9CB1FY51GH7J","wellarchitected-workload-auditor","Audits AWS Well-Architected Tool workloads for review staleness, per-pillar high-risk issue counts, milestone tracking gaps, and remediation plan completeness. Emits a deterministic verdict (STALE_REVIEW | HIGH_RISK | CONFIG_GAP | OK) per workload with per-pillar risk breakdown, milestone trend analysis, and specific remediation. Use when reviewing Well-Architected workloads, checking WA review freshness, auditing high-risk issues per pillar, validating milestone coverage, or tracking remediation plans.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwellarchitected-workload-auditor\u002FSKILL.md",[],[],{"skillNo":158,"name":159,"nameZh":6,"description":160,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":161,"categories":162,"keywords":163,"publishedAt":19},"01M7ADT59G6FTTNZYHT2QRPNNP","wellarchitected-review-operator","Operates AWS Well-Architected Tool reviews end-to-end — creates workloads, runs pillar reviews (operational excellence, security, reliability, performance, cost optimization, sustainability, Prosperity), answers questions with risk-tier rationale, generates reports and improvement plans, creates milestones, and integrates with Trusted Advisor and Well-Architected Labs. Runs deterministic pre-checks (workload existence, lens availability, pillar coverage), emits the exact update-answer \u002F get-consolidated-report CLI behind a CONFIRM gate, and verifies risk-tier transitions. Emits READY | BLOCKED | COMPLETED. Use when creating a workload, answering pillar questions, generating an improvement plan, creating a milestone, or integrating with Trusted Advisor. Triggers: well-architected tool, well architected review, workload review, pillar, operational excellence, security, reliability, performance, cost optimization, sustainability, prosperity, improvement plan, milestone, trusted advisor.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwellarchitected-review-operator\u002FSKILL.md",[],[],{"skillNo":165,"name":166,"nameZh":6,"description":167,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":168,"categories":169,"keywords":170,"publishedAt":19},"01M7ADSH5JA9QHNPPHQCD7EVA7","wafv2-web-acl-deployer","Provisions AWS WAFv2 Web ACLs with correct production defaults: scope selection (CLOUDFRONT vs REGIONAL), managed rule groups (AWSManagedRulesCommonRuleSet, SQLiRuleSet, LinuxRuleSet, WindowsRuleSet, IPReputationList, BotControlRuleSet, AmazonIpReputationList, ATP), custom rules (byte match, IP set, regex pattern, geo match, size constraint, rate-based), rule priority ordering, logging destinations (Kinesis Firehose, CloudWatch Logs, S3), resource association (ALB, API Gateway, CloudFront), rate-based rules with aggregate keys (IP, forwarded IP), Challenge and CAPTCHA actions, and account takeover prevention. Emits a READY_TO_DEPLOY checklist. Use when creating a Web ACL, attaching managed rules, configuring rate limiting, enabling WAF logging, associating a Web ACL with an ALB\u002FAPI Gateway\u002FCloudFront, or configuring CAPTCHA\u002FChallenge actions. Triggers: create WAF Web ACL, WAFv2, managed rule groups, WAF rate limiting, WAF logging, WAF CAPTCHA, WAF Challenge action, bot control, ATP managed rules.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwafv2-web-acl-deployer\u002FSKILL.md",[],[],{"skillNo":172,"name":173,"nameZh":6,"description":174,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":175,"categories":176,"keywords":177,"publishedAt":19},"01M7ADRYFAT8R73FBNCHRC573X","wafv2-web-acl-auditor","Audits AWS WAFv2 Web ACL configurations to determine whether the ACL provides effective protection: default-action posture (Allow vs Block), managed-rule-group coverage gaps, rule effectiveness (BLOCK vs COUNT, shadow\u002Fbypass rules, stale exclusions), rate-based rule correctness, logging and visibility configuration, and text-transformation bypass vectors. Emits a deterministic verdict per Web ACL. Use when reviewing a WAFv2 Web ACL for security posture, checking managed-rule coverage, validating rate-limit configuration, auditing WAF logging, or hardening a WAF before production deployment. Triggers: WAF, WAFv2, Web ACL, managed rule group, AWSManagedRules, rate-based rule, CAPTCHA, Challenge, OverrideAction, Count mode, forwarded IP, Bot Control, ATP, text transformation, WAF logging, WAF visibility, OWASP, SQLi, XSS, RFI.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwafv2-web-acl-auditor\u002FSKILL.md",[],[],{"skillNo":179,"name":180,"nameZh":6,"description":181,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":182,"categories":183,"keywords":184,"publishedAt":19},"01M7ADR8W7G72JX4KFGN03QK8Q","waf-rule-deployer","Provisions AWS WAFv2 rule sets with production defaults: Web ACL creation (CloudFront scope in us-east-1 vs Regional scope), rule groups (managed vs custom) with capacity planning in WCU (Web ACL Capacity Units, 1500 default per ACL), managed rule groups (AWSManagedRulesCommonRuleSet, AWSManagedRulesSQLiRuleSet, AWSManagedRulesAmazonIpReputationList, AWSManagedRulesLinuxRuleSet, AWSManagedRulesWindowsRuleSet, AWSManagedRulesUnixRuleSet), custom rules (byte-match, regex pattern set, geo-match, rate-based), IP set and regex pattern set resources, rule priority ordering (lowest number evaluated first), action types (allow, block, count, CAPTCHA, Challenge), bot control managed rule group, account takeover. Triggers: create WAF rule, WAF managed rule group, WAF custom rule, WAF byte match, WAF geo match, WAF rate based rule, WAF IP set, WAF regex pattern set, WAF WCU budget, WAF CloudFront ACL, WAF logging Kinesis Firehose, WAF bot control, WAF ATP, WAF label matching, WAF Challenge action, WAF CAPTCHA.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwaf-rule-deployer\u002FSKILL.md",[],[]]