Notre avis
Analyse un fichier de tâche ou de wiki pour y détecter les violations des conventions d'écriture et de structure d'un laboratoire.
Points forts
- Vérifie ligne par ligne l'ensemble des règles issues des fichiers de conventions.
- Produit un rapport détaillé et autonome, facilement exploitable par un autre agent.
- Gère correctement les exceptions (setup.md, fichiers wiki).
Limites
- Nécessite que les fichiers de conventions soient présents à un emplacement précis.
- Ne corrige pas les violations, uniquement leur signalement.
- Ne couvre pas les règles non documentées dans les fichiers de conventions.
Utilisez cette compétence avant une soumission de laboratoire ou après des modifications de tâches ou de pages wiki pour garantir la conformité aux règles.
Ne l'utilisez pas pour une révision de style libre ou pour vérifier la validité technique du contenu ; utilisez plutôt un outil de linting ou de relecture humaine.
Analyse de sécurité
SûrThe skill only reads convention files and the target file, then writes a report. It does not execute any system commands or network requests, and there is no risk of data exfiltration or destructive actions.
Aucun point d'attention détecté
Exemples
Review the file lab/tasks/required/task-2.md for convention violations.Check wiki/web-development.md against the lab conventions.name: review-via-conventions description: Review a task file for convention violations argument-hint: "<path>"
Review a single file for violations of the lab authoring conventions. The file must be under lab/tasks/ or wiki/.
Steps
- Parse
$ARGUMENTSto get the file path. Accept paths underlab/tasks/(e.g.,lab/tasks/setup.md,lab/tasks/required/task-2.md) orwiki/(e.g.,wiki/web-development.md). If the path is missing or does not point to a file under one of these directories, ask the user. - Read the target file.
- Read the convention files that apply to the target file:
- For
lab/tasks/files:instructors/context/conventions/common.md— writing conventions (4.1–4.23)instructors/context/conventions/tasks.md— task structure (Section 3) and design principles (Section 12)
- For
wiki/files:instructors/context/conventions/common.md— writing conventions (4.1–4.23)instructors/context/conventions/wiki.md— wiki file structure and section patterns
- For
- Go through the target file line by line. Check it against every convention in both files. Flag each violation with its line number.
Rules
- The convention files are the single source of truth. Check every rule they contain — do not skip any.
- Do not invent rules beyond what the convention files state.
- Be strict: flag every violation, no matter how small.
- Do not fix anything — only report.
- If a convention does not apply to the file (e.g., the file has no Docker commands), skip that category and note "Not applicable."
- For
lab/tasks/setup.md: skip task-only conventions (Section 3 template, acceptance criteria format). Apply allcommon.mdconventions.
Output format
Write the report to tmp/review-via-conventions/<repo-root-path>, where <repo-root-path> is the file's path from the repository root (e.g., tmp/review-via-conventions/lab/tasks/setup.md for lab/tasks/setup.md, tmp/review-via-conventions/wiki/web-development.md for wiki/web-development.md). Create intermediate directories if they do not exist.
The report must be self-contained so another session or agent can act on it without extra context. Structure:
- Header — file path reviewed, date, convention files used.
- Findings — grouped by convention number (e.g., "4.2. Terminal commands", "Section 3. Task document structure"). Under each group, list findings as numbered items with line numbers. If a group has no findings, write "No issues found."
- Summary — total violation count and a short overall assessment.
After writing the file, print its path in the conversation so the user can find it.
TDD Red-Green-Refactor
Testing
Skill qui guide Claude a travers le cycle TDD complet.
Audit d'Accessibilité Web
Testing
Réalise un audit d'accessibilité web complet selon les normes WCAG.
Générateur de Tests UAT
Testing
Génère des cas de test d'acceptation utilisateur structurés et complets.