kb-build

1

glenzli

An advanced execution skill that reads and executes the KB_PLAN.md plan, using a cognitive-mapping pipeline to construct concrete knowledge-base documents and perform 3D verification.

kb-init

1

glenzli

An advanced orchestration skill for initializing projects. It automatically runs kb-plan to generate a blueprint and, after requesting user confirmation, automatically launches kb-build to execute the document production cycle.

kb-update

1

glenzli

An incremental knowledge-base maintenance skill that detects document rot via fingerprint checks and performs targeted scoped updates; reference kb-build when creating tasks.

kb-plan

1

glenzli

A planning-phase skill for exploring a codebase and generating a knowledge-base construction blueprint (KB_PLAN.md). This skill performs a high-level scan of the repository during planning, distinguishing "high signal-to-noise" business logic from "boilerplate noise" and producing an actionable KB_PLAN.md build blueprint. Features include: reading only README and core configuration files under constraint (e.g., package.json / go.mod / pom.xml / docker-compose.yml), inspecting the project root and main entry skeleton files without delving into implementation details; identifying modules to include (business cores, cross-module contracts, implicit state transitions, stable public APIs) and directories to exclude (tests, node_modules/vendor, scaffold scripts, etc.). It outputs a standards-compliant KB_PLAN.md at the project root or .agent/kb/ containing overall planning, filtering declarations, and a construction task checklist with checkboxes (each item lists the target KB path, involved source code, and core focus points). After generation it stops and prompts for human confirmation to allow subsequent use of kb-build to execute documentation consolidation, improving documentation output efficiency and consistency.

moe-cr

1

glenzli

A Mixture-of-Experts (MoE) code review skill that distributes code diffs to dedicated expert prompts for parallel processing, cross-checks against a knowledge base, and generates an aggregated report with risk ratings.