Lazy-loading index. Import this file from your project's root CLAUDE.md
or CLAUDE.local.md (e.g. @.claude/loading.md) so Claude sees the rules
below every turn. The rules then tell Claude which other files to pull in
on demand — keeping the per-turn context small.
When to read: before creating a commit, opening or updating an MR, naming a new branch, or pushing a branch.
Why: defines the team's --author flag, MR checklist entries, Linear
linkage, GitLab glab usage, branch-prefix rule, the post-push MR
consistency check (launch the mr-consistency subagent after every
push), and the tezosx release backport rule (cherry-pick MR to
tezosx-athena).
When to read: before adding a new test — unit, tezt, or
property-based.
Why: requires an evaluation phase to detect redundancy first — whether the coverage already exists, whether to extend an existing test instead of writing a new one, and whether several tests already exercise the same code path.
When to read: when choosing which Claude model or effort level to
use — for a session (/model), a new subagent definition, a workflow
stage, or an API-calling script.
Why: maps task types (protocol work, review, MR text, mechanical sweeps) to models and effort levels, with cost figures and rules of thumb.
When to read: when you are about to draft or update an MR description.
Why: the required template — structure (What / Why / How / Manually testing the MR / Checklist) and the <LINEAR_ISSUE_URL> placeholder that
must be substituted.
When to read: during onboarding — only if the project-root
CLAUDE.local.md is missing and the user needs to create it.
Why: template for personal overrides (CLAUDE_COMMIT_AUTHOR,
CLAUDE_BRANCH_PREFIX).
When to read: never on demand — Claude Code auto-loads this file at session start for plugins and permissions.
When to read: if a teammate asks what this .claude/ stash is, how to
install it, or how its loading model works. Not needed for day-to-day
tasks.