What problem does it solve? AI agents often start coding without understanding a project's conventions, past decisions, open questions, or in-flight plans, causing regressions and repeated mistakes. This Skill enforces a strict read-only onboarding protocol that loads the project's full institutional memory before any work begins. ## Core Features & Use Cases - Structured Memory Loading: Walks the .ai-memory/ folder recursively, covering CODE RED rules, user preferences, institutional knowledge indexes, pending plans, issues, and CI/CD failure history. - Git Context Inspection: Analyzes the last 10 commits via git log -n 10 --stat to understand recent file changes and architectural intent. - Ambiguity Handling: Surfaces open questions from ambiguous-questions/ folders and treats resolved ones as binding decisions, refusing to guess past blockers. - Tiered Read-Only Tooling: Uses cached Python readers, GitMap CLI, or native file tools with a strict ban on repository writes during reading. - Use Case: Before modifying a large spec-driven repository, run this protocol so the agent can cite the exact file behind every rule it enforces and list all pending plans from memory. ## Quick Start Ask the agent to run the read memory protocol and load the project context before starting any code changes.
