implementer
01Agent Claude Code
Implements a planned story in an isolated context. Invoked by /ks-execute.
Agent Claude Code
Implements a planned story in an isolated context. Invoked by /ks-execute.
Agent Claude Code
Anti-hallucination review of the implementer's work, fresh context, read-only. Invoked by /ks-review.
Agent Claude Code
Reviews the story breakdown against the PRD perimeter, fresh context, read-only. Invoked by /ks-stories-review.
Command Claude Code
Set the technical HOW — stack, patterns, conventions, design.
Command Claude Code
Capture and structure the global design system into docs/design-system.md. Doesn't generate visuals — it records them.
Command Claude Code
Derive a story's screen from the design system. Autonomous path (the agent produces it) or brief path (an external tool produces it and the result comes back). Never freestyles outside the system.
Command Claude Code
Get a story implemented in an isolated subagent. Never codes in the main context.
Command Claude Code
Affiche le pipeline killer-saas — l'ordre des phases et la règle unique.
Command Claude Code
Chain a story's full cycle — Research → Design → Plan → Execute → Review → Ship — with two blocking human checkpoints (plan validation, ship confirmation).
Command Claude Code
Break a story into sequenced tasks, ready to execute.
Command Claude Code
Frame the kill — target SaaS, kill mode, perimeter, the WHAT and the WHY.
Command Claude Code
Explore a story's real context before planning — current code, APIs, traps.
Command Claude Code
Get a story reviewed by a fresh-context subagent. Gate before Ship. Never reviews in the context that wrote the code.
Command Claude Code
Ship the story per the project's Merge mode and Ship confirmation (AGENTS.local.md).
Command Claude Code
Show the project's pipeline state — framing docs, per-story progress, next command.
Command Claude Code
Review the story breakdown against the PRD perimeter, in a fresh-context subagent. Cheapest place to catch a bad split.
Command Claude Code
Break the PRD down into shippable, agentic-ready user stories.
Skill Claude CodeCodex
Breaks a need or a PRD down into user stories an agent can execute. Use during the Stories phase of the killer-saas pipeline.
Skill Claude CodeCodex
Analyzes existing code you didn't write — structure, conventions, patterns. Use during the Architecture and Research phases of the killer-saas pipeline, and for boilerplate onboarding.
Skill Claude CodeCodex
Detects agent hallucinations in generated code — invented APIs, plausible-but-wrong logic, drift from the plan. Preloaded in the reviewer subagent, and the single home of the review procedure.
Skill Claude CodeCodex
Reviews a user-story breakdown against the PRD perimeter — coverage gaps, graveyard leaks, technical-layer stories, untestable criteria, dependency order. Preloaded in the stories-reviewer subagent.
Command Claude Code
Create the project's AGENTS.local.md — the settings and commands the pipeline reads. Runs once, before /ks-prd.
Command Claude Code
One small story, end to end, in three contexts instead of six — research and plan fused, same implementer, same fresh-context review, same gates.
Skill Claude CodeCodex
How a story's screen is derived from the design system — what a new screen owes, what a derived one does not, and why nothing is measured outside the design-system phase. Loaded by /ks-design and /ks-design-system.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: