--- name: study-plan description: Builds a study plan for an AWS certification using the exam-focused methodology, from the results of a diagnostic mock exam (AWS Skill Builder, PDF, pasted text, screenshots/images or other websites). Produces the question log with confidence levels, service tiers cross-checked against the official exam guide, the prioritization matrix and a day-by-day plan up to the exam date. Use when the user asks for a study plan, study schedule or gap analysis of an AWS certification mock/practice exam (e.g., SAA-C03, SOA-C03, DVA-C02, CLF-C02, SAP-C02, DOP-C02). Also triggers on "plano de estudos", "analisar meu simulado". license: MIT compatibility: Requires Python 3 and internet access (official AWS exam guide). Reading results from a logged-in browser additionally requires Node.js and the Playwright MCP Bridge extension. metadata: version: "0.2.0" --- # AWS certification study plan (exam-focused) You apply the methodology in [references/methodology.md](references/methodology.md) to a diagnostic mock exam the user has already taken, and deliver a **study plan in markdown**. Read the whole methodology before starting. It is the source of truth for the rules (confidence, tiers, matrix, 24-hour rule, exam day). Scope: **AWS certifications only**. If the user asks for another vendor's certification, say this skill is focused on AWS. Write all output in the user's language (default: the language of the conversation), including the plan, even though the templates are in English. **Paths:** every path in this skill (`scripts/`, `references/`, `assets/`) is relative to the directory that contains this `SKILL.md`. Resolve them to absolute paths before running commands, because the working directory is usually the user's project. In the commands below, `` is that directory. ## Privacy rules (mandatory) - **Never** write to a file or repeat in your reply: tokens (e.g., `PLAYWRIGHT_MCP_EXTENSION_TOKEN`), cookies, URLs with session/registration parameters (`registration_id`, `product_id`, assessment IDs), the user's name taken from the page, e-mail or account ID. - The Playwright token comes **only** from an environment variable or the user's own MCP configuration. If the user pastes a token in the chat, use it in memory only for the run (`PLAYWRIGHT_MCP_EXTENSION_TOKEN=... python3 ...`), never in a file. - Raw dumps (`*.raw.txt`) and `questions.json` go in a **temporary directory** (`mktemp -d`), outside the output folder. Delete them at the end unless the user asks to keep them. - When citing the source in the plan, use only its name (e.g., "Official Practice Exam SOA-C03, Skill Builder"), never the link with IDs. ## Flow ### 1. Collect inputs Ask in a single message **only for what is missing**: - **Exam:** code or name (e.g., SOA-C03). - **Dates:** study start (default: today), exam date and time of day. - **Hours available per day.** If the user doesn't know, assume an estimate and state it in the plan. - **Diagnostic source:** link, file (PDF, image, CSV, TXT) or pasted text. More than one source is allowed. - **Output folder:** default = current directory. If the user **hasn't taken** a diagnostic yet, explain briefly the methodology and tell them to take a mock exam in **exam mode**, marking confidence (Knew/Unsure/Guess) on each question, and stop. ### 2. Ingest the diagnostic → `questions.json` Follow [references/question-sources.md](references/question-sources.md) for the given source and produce the normalized file described in [references/questions-json-format.md](references/questions-json-format.md). Shortcut for Skill Builder (text goes straight to a file, without flooding the context): ```bash T=$(mktemp -d) python3 /scripts/browser_extract.py --url "" --click-tab "Question Details" --out "$T/sb.raw.txt" python3 /scripts/parse_skillbuilder.py "$T/sb.raw.txt" --exam --out "$T/questions.json" ``` Check the question count against the platform's summary. If it doesn't match, investigate before moving on. ### 3. Read the official exam guide Find the exam guide (certification page at `aws.amazon.com/certification/` → "Exam guide" link → `docs.aws.amazon.com/aws-certification/...`) and download it: ```bash python3 /scripts/fetch_exam_guide.py "" --out-dir "$T/guide" ``` Extract: format (duration, number of questions, how many are **scored**, passing score), **domains and weights**, tasks/skills, **in-scope** and **out-of-scope** services. Don't use weights from memory: if you can't access the guide, say so in the plan. ### 4. Annotate each question Generate the question digest and fill in `domain`, `service` (main service, with a standardized name), `topic` (specific topic, e.g., "Route 53: geoproximity vs geolocation") and `note` (one-line takeaway from the explanation) in the JSON: ```bash python3 /scripts/summarize_log.py "$T/questions.json" --all-digest ``` For long exams, work in batches. Update the JSON with a Python script instead of editing question by question by hand. ### 5. Generate the tables ```bash python3 /scripts/summarize_log.py "$T/questions.json" ``` Output: summary, confidence × result, domains, question log and service dashboard. ### 6. Classify into tiers and prioritize - **Tier 1:** many questions in the log **and/or** a core service of a high-weight domain. **Tier 2:** frequent. **Tier 3:** rare or superficial. - Cross-check with the guide: **remove or flag what is out-of-scope** (even if it appeared in the mock exam) and flag services that are on neither list. - Build the matrix from section 6.2 of the methodology. - **Gap = wrong + skipped + right with Unsure/Guess.** A right answer with no confidence mark is neutral: ask the user if you want precision. ### 7. Build the daily plan Use the template [assets/study-plan.md](assets/study-plan.md). Rules: - **Include every gap**, in matrix order. Group related services on the same day and estimate hours per block within the available hours. - Each item of the day states **the specific topic** and the source questions (e.g., "Lifecycle hooks (Q62)"). - The **consolidation mock exam (exam mode)** must finish **≥ 24 h before the exam** (allows rescheduling). Include the keep/reschedule decision criterion. - **Exam day** is light review only (cheat sheet, flashcards). - **Time management:** compare the average time per question with the exam limit (duration ÷ number of questions). If the user skipped questions or went over the limit, include a per-question target and checkpoints. - Warn that **retaking the same mock exam inflates the score**. Prefer a different question bank to measure progress. - If there aren't enough days to cover the gaps, say so clearly and propose an alternative (more hours or rescheduling). Don't compress silently. - Include the comparison cheat sheet built from the mistakes and the final checklist. ### 8. Deliver - Write `study-plan-.md` to the output folder and delete the temporary directory. - Reply briefly: score vs passing score, number of gaps, the 3–5 biggest gaps, the day-by-day plan in one table and the assumptions made (e.g., hours/day). - Offer the next step: generate the study material with the **study-content** skill.