What problem does it solve? Pre-action analysis reports often mix raw source references (absolute paths, line numbers, status tags) with readable content, forcing developers to cross-check multiple files. This Skill turns a machine-readable facts.json into a self-contained markdown deliverable that any reader can consume without consulting other documents. ## Core Features & Use Cases - Three-layer rendering: Produces a single markdown file structured into strategic (2 min, project manager), tactical (15 min, implementing developer), and technical (30 min, auditor) reading layers, routed from claim kinds via Jinja2 templates. - Strict content/form separation: The deliverable never exposes absolute paths, line numbers, or [X.12]/[CONFIRME X.13] status tags; all references stay in facts.json for auditability. - Anti-regression validator: Before writing, the renderer rejects output containing stubs (TODO, [a definir]), unsupported HTML tags in Mermaid, relative pronouns after bullet labels, or visible source references, exiting with code 4 and a violation list. - Use Case: After running build_facts.py on a UC-100 pre-action analysis, run render_livrable.py to produce the final markdown deliverable handed to the developer, with an optional internal-audience variant that keeps inline source references. ## Quick Start Render the final markdown deliverable from the facts.json generated by build_facts.py using the render_livrable.py script with the default external audience.
