AI Agent Workflow Collection

Design, audit, or improve agentic systems and agent harnesses. Use for agent MVP blueprints, tool design, permission models, planning and goal loops, context compaction, memory, skills, connectors, observability, evals, safety, and provider-neutral agent architecture.

Review a codebase for architecture deepening opportunities: shallow modules, weak seams, scattered domain logic, low testability, high coupling, and AI-navigability friction. Use when the user asks for architecture improvement, refactoring opportunities, modularity, test seams, or maintainability review.

Create a single working brief before substantial implementation, audit, or migration work. Use when tasks are non-trivial, risks are real, or the session is likely to drift without one source of truth for scope, non-goals, verification, and done criteria.

Verify local or deployed web pages with a real browser using Playwright or equivalent browser automation. Use after frontend changes, visual fixes, responsive work, asset updates, forms, navigation, canvas/3D, screenshots, or when checking console errors, network failures, layout overflow, rendered images, and interaction flows.

Audit claimed delivery against the starting baseline and the current working tree. Use when a task has declared deliverables, when a long run might say 'done' before shipping real changes, or when final verification must check the actual artifact set rather than only the transcript or commits.

Run a baseline preflight before long implementation or autonomous execution. Use when a task has multiple phases, mandatory commands, or high risk of thrashing against an already-broken repo state. Deduplicate the minimum useful checks, surface red baseline conditions early, and require an explicit user decision before proceeding past known breakage.

Create, inspect, and apply DESIGN.md brand kits for brand-aware frontend work. Use when building or reviewing landing pages, frontend UI, pitch pages, redesigns, client audits, brand-matched prototypes, or researching a potential client/company where website identity, design tokens, typography, layout, components, or do/don't guidance should shape the output.

Stress-test a plan, feature, or design against the repo's existing docs, domain language, ADRs, and context files. Use when a task is ambiguous, terminology is fuzzy, architecture trade-offs matter, or the user wants alignment before implementation. Ask one question at a time, inspect the codebase when answers are already present, and sharpen the project's language as decisions crystallize.

Keep repo-local agent context accurate after code and workflow changes. Use when docs/ai, AGENTS.md, runbooks, or context entrypoints have drifted from the codebase, after major feature work, after architecture changes, or when an agent keeps rediscovering the same repo facts. Refresh durable context, split groups only when needed, and preserve compact read order.

Plan and run large Codex tasks with compact dynamic workflow packets, risk gates, integration, and verification. Use when the user explicitly asks for dynamic workflows, packets, subagents, swarm, orchestration, long-running audits, multi-track coding/research work, job or client search agent design, migrations, repo-wide reviews, or tasks where independent discovery, implementation, review, and verification tracks would reduce drift.

Research and answer evidence-sensitive questions with source, date, unit, and operand discipline. Use for web research, PDF/document QA, policy or finance facts, tables, time series, calculations from retrieved evidence, exact-answer extraction, and any task where wrong source selection, stale facts, nearby table confusion, or formatting errors would matter.

Build or improve frontend pages, apps, landing pages, pitch pages, redesigns, dashboards, and prototypes with strong hierarchy, responsive layout, real visual assets, ergonomic controls, and verification. Use when frontend quality, brand fit, product clarity, or design implementation matters.

Reduce repository-reading token waste by navigating code through symbols, entrypoints, dependencies, changed files, and blast radius instead of broad whole-tree or whole-file reading. Use on medium-to-large repos, PR reviews, architecture questions, onboarding, debugging across modules, or long-running audits.

Restructure agent-facing docs so the always-loaded context stays small and the rest is loaded on demand. Use when a repo has bloated AGENTS.md or CLAUDE-style docs, too many always-loaded notes, stale session history, oversized runbooks, or high startup token waste for coding agents.

Execute large tasks through explicit phases with acceptance criteria, required commands, and state updates. Use when work is too large for one implementation pass but too important to leave as a vague checklist. Keep phases falsifiable, end-to-end, and small enough that each one can be verified honestly before moving on.

Break a PRD, plan, or spec into independently executable vertical slices. Use when a product artifact needs to become agent-ready or human-ready implementation tickets, especially for client, product, and multi-track project workflows. Prefer thin end-to-end slices, explicit blockers, and issue-tracker-neutral output.

Audit a private skill, prompt pack, or assistant workflow before publishing it. Use when extracting public reusable method from client, product, team, or repository-specific agent instructions. Mark private vs mixed vs reusable content, rewrite the mixed layer, and gate release on decontextualization.

Bootstrap usable project context before deep implementation work. Use when entering a new or unfamiliar repo, installing agent workflows into an existing codebase, auditing stale docs, or preparing a project for repeatable long-running agent work. Detect the stack, ask the user about the project, summarize understanding, propose compact context artifacts, and update only with approval.

Build throwaway prototypes to answer a specific product, UI, state-machine, data-model, or workflow question before committing production code. Use when the user asks to prototype, explore variants, sanity-check logic, try designs, make a playable mock, or learn quickly with code that will be deleted or absorbed.

Lightweight grilling session for brainstorming, product clarification, and early design shaping. Use when the user wants fast alignment but there is not yet enough repo context to ground a heavier doc-based review. Ask one question at a time, propose a recommended answer, and keep momentum high.

Recover from failed execution or verification with a bounded three-step loop: retry, focused fix-spec, then honest handoff. Use when an agent is likely to spin on the same failure, when flaky environment issues blur the signal, or when you need a disciplined stop condition instead of endless optimistic retries.

Debug repositories with a hypothesis ledger, checked locations, and loop-breaking discipline. Use when tests fail, CI fails, behavior regresses, local tooling errors, flaky bugs appear, or an investigation risks repeated searches, repeated commands, circular hypotheses, or unverified fixes.

Triage bugs, enhancements, and backlog items through a small state machine that works with or without a formal issue tracker. Use when reviewing incoming requests, preparing work for an agent, deciding whether more info is needed, or keeping a multi-project backlog healthy.

Design role-specific or workflow-specific agent skill packs without collapsing everything into one prompt. Use when creating assistant layers for teams, products, functions, or operating roles. Define base-vs-difference layers, pack boundaries, references, triggers, and rollout order.