Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/emanzurv/quiz-me-skillWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/emanzurv/quiz-me-skill/review)<a href="https://agentmods.dev/commands/emanzurv/quiz-me-skill/review"><img src="https://agentmods.dev/badge/commands/emanzurv/quiz-me-skill/review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/emanzurv/quiz-me-skill/review"><img src="https://agentmods.dev/badge/commands/emanzurv/quiz-me-skill/review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00023 | $0.01550 |
| Opus 5 | $0.00012 | $0.00775 |
| Sonnet 5 | $0.00005 | $0.00310 |
| Haiku 4.5 | $0.00002 | $0.00155 |
Grade A, and why
review scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Re-quiz the user on code that already landed. Passing a quiz right after implementing proves recall at peak context; this proves retention.
Scope is the user's own session work, never the whole branch. Code someone else pushed to the branch is not something the user can be expected to have retained.
-
Collect every file this repo's sessions have edited, across all sessions. Session transcripts live under
~/.claude/projects/, one directory per project, named after the repo path with every non-alphanumeric character replaced by a dash:ROOT=$(git rev-parse --show-toplevel) PROJ=~/.claude/projects/$(printf '%s' "$ROOT" | sed 's/[^a-zA-Z0-9]/-/g') FILES=$(jq -r 'select(.message.content) | .message.content[]? | select(.type == "tool_use" and (.name == "Edit" or .name == "Write" or .name == "MultiEdit" or .name == "NotebookEdit")) | .input.file_path // empty' "$PROJ"/*.jsonl 2>/dev/null \ | grep "^$ROOT/" | sed "s|^$ROOT/||" | sort -u)Subagent edits sit in the same transcripts under
isSidechain: true; they count. -
Stop here if
FILESis empty. Do not fall through to the next step — an empty variable leaves a bare trailing--, which git reads as no path restriction and answers with the user's entire history, silently undoing the session scoping:[ -z "$FILES" ] && echo "no session-edited files in this repo — nothing to review"Empty means either no session work or no local transcripts; a fresh clone is indistinguishable from an idle repo here, and neither justifies quizzing on code the user did not write.
-
List the last
${1:-5}commits that are the user's and touch those files. Walk back as far as history needs —5means five surviving commits, not five candidates. Feed the paths throughxargs -0; do not interpolate$FILESinto the command, as zsh does not word-split unquoted expansions and would hand git one newline-stuffed pathspec that matches nothing:printf '%s\n' "$FILES" | tr '\n' '\0' \ | xargs -0 git log --author="$(git config user.email)" --oneline -n ${1:-5} --Both filters carry weight. Authorship alone still admits branch work that never went through a session; paths alone still admit other people's commits to files the user happens to have edited. Show the list.
-
Read
difficultyfrom.claude/quiz-me.jsonmerged over~/.claude/quiz-me.json(QUIZ_ME_DIFFICULTYoverrides both; unset reads asnormal). It sets how many hunks to pick, how many options each question carries, how tight the distractors are, and how wide the previews run — the table in the skill's Difficulty section governs here too. Then pick that many of the most substantive hunks — control flow, API shape, ordering, error handling. Skip formatting, renames, and generated files. Prefer hunks in files not quizzed before; the review log sits beside the misses log:f=~/.claude/quiz-me/"${ROOT//\//_}"; cat "$f".reviewed.md 2>/dev/nullROOTis the repo root from step 1, not$PWD— the hook derives the same filenames from the session'scwd, and a shell in a subdirectory would read and write a different project's state.Never-reviewed files first, then least recently reviewed. Repeating a file is fine once the unreviewed ones run out — say which bucket each question came from.
-
Read the current state of the code around each hunk, not just the diff. The question is about how the code works today.
-
Quiz with
AskUserQuestion, one question per hunk, cold — do not restate what the commit message said, and do not show the diff before asking. Open with a rule line:── COLD REVIEW · 5 session commits · 4 questions ──Fill every field:
headerasQ2/4 cache-style progress chips,descriptionas one grounded line per option, andpreviewshowing the code each option points at — same line count for every option, real lines only, all-or-none. Place the correct answer by the slot rule, not first: smallest line number it cites,modthe option count,+ 1. -
Grade each answer out loud: retained / fuzzy / lost. For anything not retained, explain the mechanism and append a line to the project's misses log under
~/.claude/quiz-me/— never inside the repo:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago Changed 9ffdea97a4f5
- 10d ago First seen · 114 lines · 23 tokens per session scan A 258fd32d3af1
review is a command published in the GitHub repository emanzurv/quiz-me-skill (5 stars, last pushed 2d ago), licensed MIT. It adds 23 tokens to every session and 1,550 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
explain
Provide clear explanations of code, concepts, and system behavior with educational clarity.
code-review-teach
Educational code review that teaches while reviewing. Provides constructive feedback, explains best practices, discusses trade-offs, and helps developers improve their coding skills.
concept-deep-dive
Deep dive into programming concepts with layered explanations, real-world examples from your codebase, practical applications, and interactive learning.
practice-exercise
Generate personalized coding exercises based on your codebase patterns and skill level. Includes progressive hints, solution review, and hands-on learning challenges.
workflow-tracer
Trace complete workflows and user journeys through the codebase.
business-logic-map
Extract and map all business logic, rules, and workflows from the codebase.