Broad fixes can introduce unrelated regressions and overwrite existing work. This skill traces each bug to its responsible layer and verifies a focused correction.

Structural code changes can introduce subtle regressions and expand scope. This skill defines behavior boundaries and brackets focused refactors with verification.

Complex migrations can lose data or break compatibility during staged rollouts. This skill structures transitions around preservation, verification, rollback, and mixed-version safety.

Feature work often grows beyond acceptance needs and weakens architectural ownership. Lean Build defines a narrow outcome, reuses repository patterns, and stops after focused verification.

Premature edits can hide causes and introduce new failures. This skill organizes evidence, ranks hypotheses, and identifies a credible mechanism before implementation.

Long memory files consume context and make repeated AI sessions expensive. This skill compresses markdown and text memory files while preserving technical details, code, URLs, and structure.

Claude Code users need reliable token receipts without asking the model to estimate usage. This skill documents a hook-based command that reports session tokens and estimated savings from the local log.

Pull request reviews often become too wordy for quick action. This skill formats feedback as concise, paste-ready comments with location, problem, and fix.

Experiment lifecycle decisions can be unsafe when evidence or server safeguards are incomplete. This skill reviews current results and recommends one supported action without executing mutations.

Aggregate repository profiles do not identify a safe or valuable code change by themselves. This skill tests one approved candidate against a controlled baseline.

Large configuration files and repeated context increase token usage. This skill reviews Caveman reports and applies measured changes with individual approval.

Users may forget Caveman modes, triggers, and setup options. This skill displays a compact reference card without changing mode or saving state.

Finding relevant code across an unfamiliar repository can consume time and context. This skill searches broadly and returns concise, verified file and line citations.

Unlabeled LLM traffic hides which jobs create spend. This skill discovers repository workflows, proposes durable labels, applies approved changes, and verifies attribution.

Large project memory files consume tokens every time an AI coding session starts. This skill compresses prose memory files while preserving technical details and saving a readable backup.

Verbose commits hide the reason for a change and slow review. This skill creates short Conventional Commit messages with context only when needed.

Verbose answers can waste context and slow technical work. Caveman mode compresses explanations while keeping code, warnings, and technical meaning intact.

Long coding sessions lose context when subagents return verbose results. Cavecrew guides Claude, Codex, and Claude Code toward compact investigator, builder, and reviewer agents.

Encrypted repository secrets require careful commands, keys, and environment handling. This skill guides SOPS workflows through an existing pnpm secrets wrapper.

Finishing a change requires careful verification, commits, pull-request setup, and CI follow-through. This skill manages that workflow until every check passes.