bastien-gallay

Analyse rhétorique et épistémologique d'articles, discours et textes argumentatifs. Utiliser ce skill quand l'utilisateur demande d'analyser la qualité argumentative d'un texte, d'identifier des sophismes ou biais, d'évaluer la fiabilité des sources citées, de déconstruire la logique d'un raisonnement, ou de produire une réécriture critique structurée d'un document.

Mise à jour totale ou partielle de la revue multi-axes du site (éditorial, visuel, tech, prose — plus perf en demi-axe et mesure/pilotage prototypé inactif), avec synthèse transverse vivante, vérification adversariale, reprise incrémentale et relecture des passes précédentes. Trigger : l'utilisateur tape `/revue`, demande à « rejouer la revue », « mettre à jour la synthèse », « relancer l'axe visuel/tech/éditorial/prose », ou « où en est la revue ? ».

Work one issue (or a scoped task) as an adversarial pair, not a delegation — the goal is to reduce the TRIPLE debt (technical, cognitive, intent), keeping the human the author of their own system. The human produces positions; the AI attacks them; the human revises. Test-first with a per-test Red→Green→Refactor→Reflect loop, two pauses, async gates that interrupt where it has leverage, a post-process retention pass (visual recap + novelty ledger), and emit-now/ store-later measurement hooks. Use when the user types `/pair-with-me #<n>`, asks to "pair on", "work this issue with me", "co-create / co-engage on a fix", or wants help on an issue WITHOUT being turned into a reviewer of AI output they can't defend.

Read a faceto event-storming board back to a stuck modeller and propose the missing pieces as reviewable events. Use when working in a repo that holds a `*.event-log.jsonl` (a faceto board) and the user asks to narrate the board, tell its story, find what's missing / what happens next, spot gaps or suspicious shapes, or "get me unstuck" mid session. Reads the log directly and — only on explicit approval, one at a time — appends proposals through a running `faceto serve` via `POST /comment` (server mints ids). Not for editing faceto's Rust, the model.json source, or the log file directly.

Critical, sourced counter-assessment ("contre-expertise") of a working method described by the user (articles, notes, practice README): extraction of the model into falsifiable propositions, author interview, multi-agent deep research (comparable methods, practices, cited sources), two-stage adversarial verification, rival method (challenger steelman), report with dated perishable verdicts, and a capped action plan in experimental format. Progresses incrementally by default (resumes from on-disk artifacts). Trigger: the user types /contre-expertise, asks to "audit my method", "critically review the way I work", "stress-test my practices against the state of the art", « passer ma méthode au crible », « analyse critique de ma façon de travailler », or wants their practice confronted with the state of the art without flattery.

Boucle d'édition critique pour écrits longs (articles, billets, essais). Le facilitateur (Claude) ne rédige PAS les paragraphes de prise de position — il critique, structure, propose des squelettes, et débusque les pièges récurrents (thèse molle, hedging, registre LinkedIn, re-cloisonnement, énumération qui esquive la nomination). Travail paragraphe par paragraphe : l'utilisateur écrit, Claude critique, l'utilisateur réécrit, Claude reporte dans le fichier. Trigger : l'utilisateur tape `/redaction`, demande à « écrire un article », « relire ce que je viens d'écrire », « critiquer ce paragraphe », ou veut aider à finaliser un brouillon long.

Job offer analysis, profile/role fit assessment, and tailored CV/cover letter generation. Activate this skill when the user mentions a job application, a job posting, a cover letter, or an adapted CV.

Pressure-test one roadmap feature per session and emit a 500–900-word decision report (verdict + ADR + adjacency + spec stub) using a closed technique library and a 6-label verdict palette (👍 ship · ✂️ reshape · ⏸ park · 🧬 split · 👎 kill · 🤷 defer-decision). Use when the user wants to take a feature from "headline on the roadmap" to "ready to scope or ready to kill" without doing a sprint plan. Reads project config from `.feature-torture.md` at project root or `.personal/feature-torture/config.md`.

Facilitate a structured brainstorming session with adaptive technique sequencing — MoSCoW, Impact/Effort, Six Hats, SCAMPER, and more from a hot-swappable technique library.