← ClaudeAtlas

jstack-devlisted

Evidence-driven enterprise engineering and mastery workflow for JStack projects. Use when the user invokes /j-stack-dev, /jstack-subagents, or /jstack-full-team, or asks for JStack planning, implementation, review, QA, security, release readiness, project handoff, or deliberate engineering training.
JarodFroneman/jstack · ★ 0 · AI & Automation · score 75
Install: claude install-skill JarodFroneman/jstack
# JStack JStack is an execution standard, not a claim that AI output is automatically production-grade. Produce the smallest coherent change, gather independent evidence, expose residual risk, and deny readiness whenever required evidence is missing, stale, incomplete, or failed. ## Command Authority The invoked command is the sole staffing authority: - `/j-stack-dev`: `single-lead`. Never spawn subagents. If the work no longer fits one lead, stop and recommend another command. - `/jstack-subagents`: `smart-subagents`. The user explicitly approved a right-sized team, normally two or three specialists. - `/jstack-full-team`: `full-team`. Account for all 11 roles and dispatch in controlled waves when concurrency would create noise. - `/jstack-loop`: use the `jstack-loop` skill. The loop owns persistence and convergence while one explicitly selected delivery mode owns each iteration. - `jstack-audit`: a separate read-only audit workflow. Use the `jstack-audit` skill; do not edit audited code or reinterpret the command as an implementation request. Do not silently reinterpret one command as another. Staffing changes coordination coverage, never the quality bar. The JStack MCP plans, validates, scans, records, and evaluates evidence. It does not spawn platform subagents. Use the platform multi-agent tools for real dispatch, waiting, collection, and closure. When an active loop supplies a `loopId`, execute only the current iteration in the contract's fixed delive