Claude Code: AI Agent Workflow Collection

Use when a user wants to enable or understand Claude Code's auto permission mode. Explains permissions.defaultMode "auto" honestly (a safety classifier approves routine actions and falls back to default mode if unavailable), and helps configure ~/.claude/settings.json with a vetted allow set and a deny floor.

Use when creating, reviewing, or editing a CLAUDE.md (or nested CLAUDE.md) file. Defines what belongs in project memory versus what is derivable bloat — high-value build/test commands, conventions, gotchas, and safety rules — and how to keep it lean and skimmable.

Use before starting a non-trivial coding task to gather the relevant repository context (README, docs, ADRs, related modules, recent git history) and distill it into a short working brief so the agent starts grounded instead of guessing. Triggers on "pull context", "get me up to speed", "brief before I start", "gather context".

A model-selection cost policy for Claude Code — route bulk, mechanical work to cheaper/faster models and reserve frontier models for judgment and review. Covers the cost/quality tradeoff, a decision matrix, how to switch models per task and per subagent, and orchestration patterns that minimize total cost to a correct result. Use when deciding which model to run or how to control spend.

Use when a code change needs end-to-end verification before handing it to a human — reproduce the change through the real running system, confirm the happy path plus at least one edge case, capture concrete evidence, and only then declare it ready.

Use when a task should repeat on a schedule or until a condition is met. Explains Claude Code's /loop taxonomy — interval loops (fixed cadence) vs. self-paced loops (model decides when to continue) — and how to set stop conditions and avoid runaway loops. Triggers on "loop", "run this every N minutes", "keep running until", "recurring task", "poll for status".

How to enable OpenTelemetry (OTel) export of Claude Code usage telemetry and configure organizational spend governance (budgets, alerts, spend caps). Team/enterprise-oriented. Covers the observability-vs-enforcement distinction, the shape of OTel configuration, and where to find authoritative specifics. Use when setting up team-level monitoring or cost controls for Claude Code.

How to split a large task into independent parallel lanes run by separate agents or git worktrees, partition work to avoid merge conflicts, and merge the results back into one coherent whole. Use when a task is large enough to benefit from fan-out orchestration.

Use when a user is repeatedly prompted to approve the same safe, read-only tool calls and wants to reduce friction. Analyzes recently denied or repeatedly-prompted tool calls, isolates the read-only ones, and proposes exact-match permissions.allow rules for .claude/settings.local.json. Never allowlists write, execute, network, or destructive commands.

How to define scheduled routines in Claude Code — cron-style cloud agents that run a task on a recurring cadence. Covers cadence selection, guardrails for unattended execution, self-contained task prompts, and worked examples (nightly PR triage, dependency-update check). Use when setting up recurring automated agent work.

Use when authoring, structuring, or connecting Claude Code skills — writing a SKILL.md, crafting the frontmatter description that drives routing, deciding when to split one skill into several, referencing other skills, and organizing a coherent agentic skills library. The reference for building skills that reliably trigger and compose.

JavaScript and TypeScript coding standards, conventions, and best practices. Use when writing, reviewing, or testing JS/TS code.