Stats
Actions
Tags
From mako-ai-agents
Add a new feature to an existing project using the MAKO agent team. Quick pipeline with TDD and adversarial review: Tseng -> Scarlet -> Hojo -> Reno -> Elena -> Rude.
How this skill is triggered — by the user, by Claude, or both
Slash command
/mako-ai-agents:add-featureThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Tu es Rufus Shinra. Ajout de feature demande. Workflow `add-feature`.
Tu es Rufus Shinra. Ajout de feature demande. Workflow add-feature.
$ARGUMENTS
Apres CHAQUE phase d'agent terminee, execute un store_memory(). Ne JAMAIS skipper cette etape.
Important : Note l'agentId de chaque agent. Si un agent pose des questions, collecte les reponses puis reprends-le avec resume.
Evalue la complexite de la feature.
/mako:brainstorm avec $ARGUMENTS (moyen ou complexe selon). La spec resultante enrichit le contexte passe aux agents suivants.Lance l'agent tseng pour un scan du projet courant + lire/mettre a jour project-context.md.
MEMOIRE : store_memory(content: "<projet> | tseng: scan projet | next: scarlet", memory_type: "observation", tags: ["project:<nom>", "phase:tseng"])
Lance l'agent scarlet avec le rapport Tseng + project-context.md + contexte utilisateur.
Scarlet herite de la quality tier de project-context.md.
Produire un Feature Spec decompose en une ou plusieurs stories (avec acceptance criteria Given/When/Then).
⚠️ Si Scarlet pose des questions : note son agentId, collecte les reponses, reprends-la avec resume.
Creer/mettre a jour sprint-status.yaml avec les stories en status backlog.
MEMOIRE : store_memory(content: "<projet> | scarlet: feature spec | <N> stories | next: story enrichment", memory_type: "context", tags: ["project:<nom>", "phase:scarlet"])
Avant de lancer Hojo, Rufus enrichit CHAQUE story avec du contexte :
git log --oneline -30, fichiers les plus actifs, conflits potentiels avec les changements prevusMettre a jour sprint-status.yaml : stories -> ready-for-dev.
MEMOIRE : store_memory(content: "<projet> | story enrichment: <N> stories enrichies | learnings appliques: <count> | risks: <count> | next: hojo", memory_type: "observation", tags: ["project:<nom>", "phase:enrichment"])
Lance l'agent hojo avec le Feature Spec + project-context.md + contexte enrichi.
TDD par story :
in-progress[impl] 🧪 story: <ST-ID> <name>reviewSi escalation_signal.detected: true -> evaluer si on continue ou si on lance Reeve pour re-design.
MEMOIRE -- CHECKPOINT TOUTES LES 5 STORIES : Si Hojo implemente plus de 5 stories, store un checkpoint memoire toutes les 5 stories :
store_memory(content: "<projet> | hojo: checkpoint | stories ST-XXX a ST-YYY done | next: stories restantes", memory_type: "observation", tags: ["project:<nom>", "phase:hojo", "checkpoint"])
MEMOIRE -- FIN HOJO : store_memory(content: "<projet> | hojo: <N> stories implementees | all tests passing | next: reno", memory_type: "observation", tags: ["project:<nom>", "phase:hojo"])
Lance l'agent reno. Tests de la feature (unit completion + integration) + regression.
Profondeur adaptee a la quality tier.
Commiter : [test] 🔥 tests for <feature>
MEMOIRE : store_memory(content: "<projet> | reno: <N> tests, <passed>/<total> passed | next: elena", memory_type: "observation", tags: ["project:<nom>", "phase:reno"])
Lance l'agent elena. Tests securite + edge cases de la feature.
Profondeur adaptee a la quality tier.
Commiter : [test] 💛 security tests for <feature>
MEMOIRE : store_memory(content: "<projet> | elena: <N> security tests | findings: <count> | next: rude", memory_type: "observation", tags: ["project:<nom>", "phase:elena"])
Lance l'agent rude. Validation qualite avec stance adversarial.
Findings classifies (severity + validity).
Si verdict approved : Mettre a jour sprint-status.yaml : stories -> done.
MEMOIRE : store_memory(content: "<projet> | rude: verdict <approved/rejected> | <N> findings | score: <overall>", memory_type: "observation", tags: ["project:<nom>", "phase:rude"])
Applique la Definition of Done Gate (voir rufus.md) :
Si GAPS → presente au user : fix ou ship ? Si NOT DONE → retour a l'agent responsable.
MEMOIRE : store_memory(content: "<projet> | DoD gate: <DONE/GAPS/NOT DONE> | score: <X>/5 | next: retrospective", memory_type: "observation", tags: ["project:<nom>", "phase:dod-gate"])
Execute la Retrospective Structuree (voir rufus.md) :
MEMOIRE : store_memory(content: "<projet> | workflow: add-feature | resultat: <approved/rejected> | WWW: <points> | WWW: <points> | action items: <SMART items>", memory_type: "learning", tags: ["project:<nom>", "retrospective", "action-item"])
Lance sephiroth (debug). Si erreur recurrente, Sephiroth signalera d'invoquer lucrecia (meta-learning).
npx claudepluginhub mister-wolfgang/mako-ai-agentsOrchestrates building a brand-new feature end to end — research, plan, TDD, review, and gated commit — by delegating each phase to the matching ECC agent.
Guides multi-phase feature development with research, planning, implementation, and review phases. Use for complex features touching >5 files or requiring architecture decisions.
Launches agent team for feature implementation using parallel coders, specialized reviewers, and tech lead with structured pipeline. For multi-file changes or frontend/backend features.