grimoire-forge

Creation, edition et validation d'artefacts Grimoire via BMB. Use when: create agent, edit agent, validate agent, create module, edit module, validate module, create workflow, edit workflow, validate workflow, builder factory, artifact design.

Compression lossless de documents optimisée LLM. Produit des distillats hyper-compréssés preservant chaque fait, décision et contrainte. Use when: distill documents, compress docs, create distillate, optimize context, reduce tokens, document compression, context optimization, LLM-optimized docs.

Audit qualité d'un SKILL.md ou d'un hook Grimoire avec grille de notation 100 points et seuils de blocage. Use when: review d'une skill, audit qualité skill, scoring SKILL.md, valider une skill avant publication, gate de qualité skill-forge, vérifier hook security, contrôler artefact .github/skills ou .github/hooks, before promoting dynamic skill, check skill description quality.

Création gardée de skills et de hooks Grimoire avec template, normes, self-test et gate qualité bloquant. Use when: créer une skill, nouvelle skill, créer un hook, nouveau hook, scaffolder une skill, design skill, design hook, composer un skill, builder skill, write a SKILL.md, écrire SKILL.md, promote dynamic skill to permanent, create lifecycle hook, hook security guardrails.

TRIGGERS for Workflow 10 (Release Architecture Version) — invoke this skill FIRST, do not plan or ask clarifying questions, when user says any of: 'release my architecture', 'release architecture', 'release architecture version', 'publish architecture', 'ship architecture', 'tag architecture version', 'freeze architecture', 'bump architecture version', 'finalize architecture' — these route here, NOT to architecture-docs-export (which only produces Word .docx files). Also use this skill for: creating/updating/maintaining ARCHITECTURE.md, generating Mermaid / C4 diagrams (Workflow 8), migrating to docs/ multi-file layout (Workflow 9), validating/auditing architecture (BIAN, META, standards), answering questions about documented components, data structures, integrations, security, performance, deployment, technology stack, or architectural decisions.

TRIGGERS for Workflow 10 (Release Architecture Version) — invoke this skill FIRST, do not plan or ask clarifying questions, when user says any of: 'release my architecture', 'release architecture', 'release architecture version', 'publish architecture', 'ship architecture', 'tag architecture version', 'freeze architecture', 'bump architecture version', 'finalize architecture' — these route here, NOT to architecture-docs-export (which only produces Word .docx files). Also use this skill for: creating/updating/maintaining ARCHITECTURE.md, generating Mermaid / C4 diagrams (Workflow 8), migrating to docs/ multi-file layout (Workflow 9), validating/auditing architecture (BIAN, META, standards), answering questions about documented components, data structures, integrations, security, performance, deployment, technology stack, or architectural decisions.

Direction artistique, production et review des assets 2D du Grimoire Game. Use when: asset 2D, pixel art, sprite, FX, room kit, style guide, palette, visual asset review, art direction, polish visuel.

Assess Grimoire project resilience and anti-fragility. Use when: antifragile score, resilience check, how robust is the project, system health, failure analysis, early warning, stress test, project resilience, fragility assessment. Combines antifragile scoring, early warning system, and failure museum analysis.

Review d'architecture multi-dimension. Évalue la qualité structurelle, les couplages, la dette technique, et la conformité aux ADR. Use when: architecture review, structure review, coupling analysis, tech debt, dependency check, modularity, evaluate architecture, ADR conformity.

Brainstorming structuré pour features, architecture, et solutions. Use when: brainstorm, generate ideas, explore approaches, propose solutions, architecture options, design alternatives, creative exploration, idea generation.

Generate a structured CHANGELOG from git history and session chain. Use when: changelog, release notes, what changed, version bump, prepare release, summarize changes, diff since last release, release prep. Combines git log analysis with session chain context for rich changelogs.

Review de code multi-couches avec triage structuré. Lance 3 reviewers parallèles (Adversarial, Edge Case Hunter, Acceptance Auditor) puis trie les findings. Use when: code review, review this code, review my changes, review staged, review branch diff, PR review, review avant merge.

Architecture, UX, stories et execution BMM. Use when: create architecture, create UX design, epics and stories, sprint planning, sprint status, implementation readiness, correct course, create story, dev story, retrospective, quick spec, quick dev, delivery flow, execution pack.

Génère un prompt de dispatch structuré et validé avant chaque runSubagent. Use when: dispatch to sub-agent, prompt engineering for handoff, multi-agent task, complex delegation, quality-gated dispatch.

Documentation, review editoriale et maintenance du corpus. Use when: explain concept, write document, validate document, update standards, mermaid, document project, editorial review, prose review, structure review, index docs, shard doc, documentation engineering.

Run Grimoire dream consolidation and insight pipeline. Use when: dream mode, consolidate learnings, off-session insights, what did we learn, cross-domain patterns, generate insights, emerging patterns, stigmergy signals, dream journal. Runs dream analysis, deposits stigmergy signals, and optionally seeds incubator with actionable ideas.

Analyse exhaustive d'edge cases par path enumeration. Orthogonal à la review adversariale — method-driven, pas attitude-driven. Use when: edge case analysis, edge cases, boundary conditions, path tracing, unhandled paths, missing guards, exhaustive review, code safety, defensive coding.

Exécution méthodique de plans d'implémentation. Use when: execute plan, run plan, follow plan, implement from plan, step by step execution, plan runner, plan executor.

Gestion active de la friction conversationnelle via QEC et budget de questions. Use when: too many questions, batching, friction budget, reduce interruptions, smooth execution, autonomous flow. Converts scattered clarifications into minimal high-value question batches.

Run a full Grimoire project health check. Use when: project audit, health check, sanity check, validate project, check project health, diagnose issues, preflight, harmony check, quality score, early warning. Combines 5 diagnostic tools into one unified report.

Réponse à incident sur code projet et infra (production, CI, build, runtime utilisateur — PAS les workflows agentiques internes Grimoire). Use when: incident production, panne, régression sur code, build/CI cassé, hotfix urgent, rollback déploiement, post-mortem code, ça marche plus en prod. Pour échec d'un workflow agentique Grimoire interne, utiliser grimoire-self-heal.

Run the Grimoire innovation pipeline from idea to prototype. Use when: new idea, innovation, brainstorm to code, incubator, R&D cycle, experiment, explore alternative, what-if scenario, quantum branch, parallel timelines, prototype, seed idea. Combines incubator lifecycle, R&D engine, and quantum branching.

Routage d'intention avec scoring de confiance pour sélectionner skill/agent/profil. Use when: intent classification, routing decision, dispatch strategy, choose best skill, ambiguous request, confidence-based routing. Produces deterministic routing decisions with explicit confidence and fallback path.

Système d'apprentissage opérationnel cross-session. Use when: log learning, save insight, operational discovery, what did we learn, remember this, cross-session memory, compound learnings.

Deep audit of Grimoire memory system. Use when: memory audit, check memory, memory issues, stale memories, memory cleanup, memory health, contradictions in memory, duplicate learnings, memory freshness, session chain review. Combines memory-lint, freshness scoring, and session chain analysis.

Profiling et optimisation de performance. Diagnostic systématique des goulots d'étranglement. Use when: slow, performance, bottleneck, optimize, profiling, latency, memory usage, speed up, too slow, benchmark.

Développement du Pixel Observatory — game engine, sprites procéduraux, timeline engine. Use when: pixel observatory, game engine, sprite system, office view, agent animation, timeline scrubber, pixel art, gamification, observatory v2.

Pre-push validation checklist for Grimoire Kit. Use when: before pushing, pre-push check, validate before commit, pre-commit, ready to push, CI check, final validation, run all checks. Runs tests, lint, harmony check, and preflight in sequence.

Recherche, cadrage produit et PRD BMM. Use when: product brief, create PRD, edit PRD, validate PRD, market research, domain research, technical research, generate project context, product discovery, cadrage produit.

Explore Grimoire project structure and hidden patterns. Use when: project archeology, hidden patterns, dark matter, desire paths, project graph, orphan files, unused agents, dead code, bus factor, tribal knowledge, who knows what, implicit assumptions. Combines project-graph, dark-matter, and desire-paths analysis.

Refactoring systématique avec filet de sécurité. Use when: refactor, refactoring, restructure, extract, simplify, reduce complexity, clean up, code smell, duplication, split module, rename.

Safety guardrails: careful mode, freeze zones, guard mode. Use when: careful mode, safety mode, restrict edits, freeze directory, guard, protect files, high-risk changes, production changes.

Security review systématique pour code et configuration. Use when: security review, audit sécu, vulnérabilité, OWASP, secrets, injection, XSS, SSRF, security check, audit sécurité, code sécurisé.

Diagnose and repair Grimoire framework workflow failures (agent runs, framework tools, internal pipelines — NOT production code). Use when: Grimoire workflow failed, agent execution broke, framework tool error, immune scan, self-healing, failure museum entry, what went wrong in the framework. For production code regressions or build/CI breakage use grimoire-incident-response instead.

Bootstrap a new Grimoire session with full context. Use when: start session, new session, session bootstrap, load context, resume work, what happened last time, session recap, catch me up, continue where we left off. Loads session chain, checks project health, restores working context.

Canon de structure du depot, hierarchie de recherche et choix de landing zone durable. Use when: structure projet, hierarchie du repo, ou chercher, ou mettre un fichier, source de verite, landing zone, gouvernance memoire, report vers grimoire-kit.

Développement piloté par sub-agents avec review en 2 étapes. Use when: executing implementation plans, multi-task implementation, subagent development, plan execution, task dispatch.

Systematic debugging workflow with 4 phases. Use when: bug found, test failure, unexpected behavior, error diagnosis, debugging, root cause analysis, fix not working.

Test-Driven Development workflow for Grimoire projects. Use when: implementing features, fixing bugs, writing tests, TDD, red-green-refactor, test first.

Strategie de test, ATDD et architecture QA/TEA. Use when: test design, test review, ATDD, NFR testing, CI testing, testing framework, teach me testing, test architecture, automation strategy, qa architecture.