lehre

Use after lehre-gauge to turn a violation set into an ordered remediation plan behind an explicit human approval gate, written to LEHRE_BRIEF.md. Trigger on 'plan the cleanup', 'what order should we fix these in', 'write the remediation brief', or 'lehre brief'. Refuses to run without a gauge result, and refuses to overwrite a brief written by another pipeline.

Use after lehre-preflight (brownfield) or lehre-decompose (greenfield) to research a code style, pattern and architecture doctrine from external authority plus real repository evidence, and write it to .lehre/ruleset.json as machine-checkable rules. Trigger on 'establish our code standards', 'what rules should this project follow', 'codify our architecture', 'research best practice for this stack', or 'lehre codify'. Every rule must cite an authority; a rule nobody can justify is refused. Not for mining the business or domain logic the code already implements — calculations, validations, state transitions — use befund-extract-rules for that. Not for a standalone, human-readable per-domain handbook (code, design, testing, documentation) recording what a project already does, one rule per dimension — use cupertino:cupertino-handbook-draft for that. This skill governs how code is written, not what the code decides.

Use to build a greenfield unit, or to apply exactly one approved phase of LEHRE_BRIEF.md, with the doctrine's blocking rules enforced at the tool-call layer throughout. Trigger on 'build the next unit', 'implement phase 1', 'apply the approved fixes', 'make this conform', or 'lehre conform'. Refuses to act on an unapproved brief, and never marks its own work validated.

Use on a blank page, after lehre-preflight reports greenfield, to turn a stated project intent into build units, the seams between them, and an enforced build order. Trigger on 'start this project properly', 'initialize this repo', 'break this down into pieces and connect them', 'what should I build first', or 'lehre decompose'. Produces units whose order is denied at the tool-call layer, not a plan anyone can skip.

Use to sweep a codebase against the doctrine in .lehre/ruleset.json and report every violation with file:line, after lehre-codify has written rules. Trigger on 'check this code against our standards', 'where do we violate our own architecture', 'audit for antipatterns', 'gauge this repo', or 'lehre gauge'. Changes no source file — reports violations, never fixes them, and records each sweep under .lehre/gauge/.

Use once a unit or phase is validated, to make its rules survive without the plugin: emit a CI check that runs the doctrine with no agent in the loop, and write behaviour tests around the code that was changed to conform. Covers CI pinning of the ruleset, durable enforcement once the plugin is uninstalled, and regression tests for conformance edits. Two phases — the durable rule check, then the regression net.

Use FIRST, before any other lehre skill, to determine whether this repository is a blank page or an existing tree, report what doctrine it already declares, and say plainly what lehre can and cannot enforce here. Trigger on 'set up code standards', 'enforce our architecture', 'start a new project properly', 'lehre preflight', or as the automatic first step of any lehre pipeline. Read-only — never writes a ruleset.

Use to report where a lehre run currently stands — which units are validated, blocked or ready, how many rules are in force, and what is denied at write time right now. Trigger on 'lehre status', 'where are we', 'what's blocked', 'what can I build next', or when resuming work on a repository that already has a .lehre directory. Read-only.

Use after lehre-conform to independently check that a unit satisfies its rules AND actually does what the project intent said it would, then close it so dependent units become buildable. Trigger on 'validate this unit', 'is this actually done', 'check the work before moving on', or 'lehre validate'. The ONLY skill that writes a unit done-marker — nothing else may certify a unit complete.