← ClaudeAtlas

plan-do-check-actlisted

Execute evidence-based experimentation cycles to test hypotheses and measure the results of a change.
Git-Fg/taches-principled · ★ 0 · AI & Automation · score 73
Install: claude install-skill Git-Fg/taches-principled
## Runtime persistence `.principled/` (in cwd) is the natural runtime emplacement for principled-related artifacts. At intake, read whatever is there if any — prior context may inform this work. When this skill produces durable artifacts, write them to `.principled/` too. Skip if absent. ## Routing Guidance - IMMEDIATELY when solving problems where outcomes need measurement — BEFORE concluding, standardizing, or shipping. - FIRST after a failed fix — validate the root cause was correct before closing the issue. - DO NOT use for debugging (use diagnose instead), for code style decisions (use refine in POLISH mode), or for architectural design (use kaizen or ideation). - CONTRAST with kaizen: kaizen prevents bad patterns from entering the codebase (proactive constraint); PDCA tests whether a change actually improves things (evidence-based validation). Use kaizen when writing new code; use PDCA when deciding whether to standardize or generalize an existing change. ## Decision Router IF testing a hypothesis to solve a problem or improve a process → start a PDCA cycle with a clear success criterion IF an experiment produced unexpected results → begin a new cycle with an adjusted hypothesis and measurement plan IF an experiment succeeded → standardize the change in the Act phase and close the cycle IF an experiment partially succeeded → standardize what worked and start a new cycle for what did not IF stuck after three cycles on the same problem → revisit the root cause analys