What problem does it solve? AI-generated product management work is often distrusted because it produces confident, evidence-free claims, silently makes prioritization decisions, and replaces the PM's judgment with unverifiable synthesis. This Skill defines three shared doctrines that keep every PM agent's output grounded, gated, and reviewable. ## Core Features & Use Cases - Corpus-grounding contract: Requires every load-bearing claim to cite the team's own evidence (intake items, support tickets, analytics, interview notes) or be explicitly flagged as an untested assumption, with fabrication treated as a cardinal sin. - Decision gates: Ensures prioritization, strategy, roadmap state, and triage outcomes are emitted as marked, reversible, explainable proposals that a human accepts, never auto-applied decisions. - Compare-don't-replace synthesis: Structures research synthesis as a traceable second opinion with outliers and disconfirming signal surfaced, inviting the PM's own read instead of foreclosing it. - Use Case: When asked whether to prioritize a bulk-import feature, the agent cites the six intake items and 22 support tickets behind the demand, shows its RICE math with confidence gaps, and proposes a ranking the human can accept or override. ## Quick Start Load the pm-agent-doctrine skill before any PM authoring or review pass and check each output against the grounded, gated, and compared quick-check questions.
