oco

Read many files in isolation and produce a compact evidence summary. Use when broad codebase exploration would bloat the main context.

Review a proposed change for consistency, correctness, and completeness. Use after code changes to verify quality before completion.

Inspect refactor scope, hidden impact, and risky omissions. Use during or after refactoring to catch missed references and breaking changes.

Orchestrated coding workflow — the unified entry point for all OCO capabilities. Use when the user types /oco followed by any request, or when a task would benefit from structured orchestration combining code intelligence, impact analysis, verification, and multi-tool coordination.

Structured codebase exploration with OCO-backed code intelligence. Auto-activates when the user asks to explore, understand, explain how a module works, a feature, an architecture, a data flow, or asks "how does this work", "where is", "show me", "what is this module". Enforces: compact summary before any action, selective reading (no directory dumps), explicit confidence level.

Systematic evidence-first bug investigation without a full stacktrace. Auto-activates when the user reports a bug, broken behavior, unexpected results, a regression, something not working, or a problem without a clear error message. Enforces strict workflow: understand symptom → narrow scope → gather evidence → reproduce → root cause analysis → fix ONLY after proof. Never guess at fixes. Also auto-activates after 2 failed attempts to fix the same problem.

Structured refactoring with impact analysis, staged changes, and verification. Auto-activates when the user asks to refactor, rename, restructure, extract, move, reorganize, decouple, or split code. Enforces a strict workflow: impact analysis before any change, staged modifications (implementation → consumers → tests → docs), full verification after each stage, subagent review if >10 files impacted. MANDATORY for all refactoring — never rename/move without this skill.

Stack trace and runtime error analysis to identify root cause. Auto-activates when the user pastes a stacktrace, traceback, panic, exception, crash, or error with line number. Also detects runtime errors in command output (build, test, run). Enforces workflow: parse trace → map to code → inspect suspect regions → ranked hypotheses → verify before fixing. Never propose a fix without reading the failing code.

Mandatory verification suite after any code change. Auto-activates after every source file modification, even trivial ones (one-liner). Detects project type (Cargo.toml, package.json, pyproject.toml, go.mod, .csproj) and runs in order: build → types → lint → tests. Produces a PASS/FAIL/PARTIAL verdict. NON-NEGOTIABLE: never consider a change complete without running this skill. Also activates when the user asks to verify, test, validate, or check their changes.

11 hooks across 10 events