Study guide skill

This skill is a navigator and quizzer over the repo's existing content (outline.md, domain-*/README.md, domain-*/examples/*.py, quiz_bank.md, exercises/). It never writes new study material — no new questions, no new example scripts, no rewritten theory notes. If a gap in the content is found, say so and suggest the user open an issue/PR upstream; don't patch it inline. Passive authoring doesn't build retention here — quizzing and doing does.

Progress file

State lives in study-progress.md at the repo root, created on first use if missing (and added to .gitignore — this file is per-learner, never committed). Structure:

# Study Progress — CCAR-F

Started: <date>

## Domain 1 — Agentic Architecture & Orchestration (27%)
- [ ] README read
- [ ] 01_agentic_loop.py run
- [ ] 02_coordinator_subagent.py run
... (one line per example file, in numeric order)

Quiz attempts: Q1✅ Q3❌ Q7✅ ...
Weak task statements: 1.5, 1.7   <!-- any T.S. with an unresolved ❌ -->

## Domain 2 ...
## Domain 3 ...
## Domain 4 ...
## Domain 5 ...

## Exercises
- [ ] 01_multi_tool_agent_escalation
- [ ] 02_team_workflow_config
- [ ] 03_extraction_pipeline
- [ ] 04_research_pipeline

## Session Log
- <date>: <one-line summary of what happened this session>

A "weak" task statement is one where the most recent attempt on any question tagged to it was ❌. Answering it correctly on a later attempt clears it.

Commands

status (default, no args) — read study-progress.md (initialize it from the template above if it doesn't exist, scanning each domain-*/examples/ dir to list actual filenames). Print a compact table: per domain, README read?, examples run X/Y, quiz score, weak T.S. list. Then hand off to next.

next — recommend exactly one concrete next action, in this priority order:

1. Any weak task statement → re-quiz it (few questions, fast feedback loop).
2. The next unread domain README, in weight order: D1 → D3 → D4 → D2 → D5 (see root README.md for why this order).
3. The next un-run example in the current domain.
4. If all examples in a domain are run and no quiz attempted yet for it → quiz the whole domain.
5. Once all 5 domains are read + quizzed → the next un-started exercise in exercises/ (numeric order). Give the recommendation as one sentence plus the reason, not a menu.

quiz <domain N | T.S. x.y> — pull the matching questions from quiz_bank.md (match on the ### T.S. x.y heading or ## Domain N heading). Prioritize, in order: previously-missed questions, then unattempted ones. Ask one question at a time, full text, options included, and wait for the user's answer before revealing anything. Do not print the <details> answer key up front. After each answer: say correct/incorrect, give the explanation from the answer key, then record the result in study-progress.md (append to Quiz attempts, update weak T.S. list) before moving to the next question. Stop after all matched questions are covered or the user says stop.

run <example-path-or-number> — actually execute it (python3 domain-N-.../examples/NN_name.py) via Bash. If it exits 0, mark that line done in study-progress.md. If it fails, report the error — don't silently mark it done, don't fix the repo's example code as a side effect (that's a separate, explicit ask).

Notes

• Task statement numbers in quiz_bank.md (T.S. x.y) match outline.md and each domain's README.md exactly — cross-reference by that tag.
• Never reveal quiz answers except one at a time, after the user has answered.
• Keep status output short — a table and one recommended next step, not a restatement of the whole repo README.