Skill Claude CodeCodex
Use when answering questions about function callers/callees, code dependencies, impact analysis, or 'who calls X' — structural search over Graphify code graph. Activates Tier-2 routing.
Skill Claude CodeCodex
Use when answering questions about past decisions, prior sessions, or 'why' questions — semantic search over Qdrant session history. Activates Tier-1 routing.
Skill Claude CodeCodex
Use when starting a new feature, system, or architectural change that is already chosen and clear enough to design — to produce approved design options and a saved spec, before any code is written, implementation skills are invoked, or project scaffolding begins. A highly detailed task description is INPUT to this…
Skill Claude CodeCodex
Use when context compaction warnings appear, after 8+ major tool sequences, after 3+ large file reads plus 2+ review cycles, when a sprint task closes mid-session with more work pending, or when the user asks for a handoff prompt or to start a new session (in any language).
Skill Claude CodeCodex
Use when facing 2+ independent tasks or failures with no shared state - to investigate or fix them concurrently without blocking each other.
Skill Claude CodeCodex
Use when executing a written implementation plan inline after the user chooses not to use subagent-driven-development.
Skill Claude CodeCodex
Use when implementation is complete and all code review is approved - to commit staged changes, integrate the branch, and update the Second Brain.
Skill Claude CodeCodex
Use when the user says '/initSumela', 'init SumelaOS', 'kur SumelaOS', or 'setup SumelaOS' in an existing project — auto-detects tech stack, architecture, and conventions, then generates AGENTS.md, rules, wiki, and IDE pointers.
Skill Claude CodeCodex
Use when a developer joins a project that ALREADY has SumelaOS committed — '/onboardSumela', 'onboard SumelaOS', 'join this project', 'set up my clone' (or the equivalent in any language). Wires git hooks, sets the per-developer interaction language + domains, and offers the optional memory runtime — WITHOUT…
Skill Claude CodeCodex
Use when performance requirements exist, a regression is suspected, or profiling reveals bottlenecks that need fixing.
Skill Claude CodeCodex
Use when receiving code review feedback from a subagent reviewer or human partner - before implementing any suggested changes.
Skill Claude CodeCodex
Use when completing a task, implementing a major feature, or before merging - to catch issues before they cascade into committed history.
Skill Claude CodeCodex
Use whenever the task involves planning, writing, or reviewing code — load at spec/plan time (brainstorming and writing-plans count), not first-code time — and unconditionally when it touches user input, forms, APIs, database queries, authentication, authorization, passwords, file uploads, permissions, secrets, CORS…
Skill Claude CodeCodex
Use after every user turn to capture correction, confirmation, decision, friction, challenge, resolution, or preference signals (resolution = bugs the agent fixes itself; preference = standing user instructions); or when the user invokes /evolve, says 'evolve', 'review pending improvements', or asks to review captured…
Skill Claude CodeCodex
Use when preparing to deploy to production — to run pre-launch checks, plan a staged rollout, set up monitoring, and document a rollback strategy.
Skill Claude CodeCodex
Use when executing an implementation plan with independent tasks in the current session — preferred over executing-plans for quality and context isolation.
Skill Claude CodeCodex
Use when encountering a bug, test failure, or unexpected behavior — before attempting any fix.
Skill Claude CodeCodex
Use when about to start implementation of any feature or bug fix — before writing any production code.
Skill Claude CodeCodex
Use when starting any implementation task that requires an isolated workspace — before creating branches or writing any code.
Skill Claude CodeCodex
Use when starting a session, ingesting a raw source, finishing a branch, answering a question that may reference prior work, or facing an information gap during reasoning (entity definitions, past decisions, call graphs, prior session context).
Skill Claude CodeCodex
Use at the start of every conversation and before generating any response — including clarifying questions — to dispatch the right skills, capture signals, and check for information gaps.
Skill Claude CodeCodex
Use when about to claim work is complete, fixed, or passing - before task completion, moving to the next task, code review, or committing any changes.
Skill Claude CodeCodex
Use when starting implementation from an approved spec or design — before any code is written or worktree is created.
Skill Claude CodeCodex
Use when creating new skills, editing existing skills, or verifying skills work before deployment.