Prompts that actually work

Claude is one of the strongest AI assistants for coding — especially for larger codebases, debugging, and code review. These prompts are tested to get the best results from Claude 3.5 Sonnet and Claude 3 Opus for software development tasks.

Claude excels at structured financial analysis, scenario modeling, and breaking down complex financial concepts. These prompts are built for personal budgeting, retirement planning, debt payoff strategies, and investment education — tested on Claude 3.5 Sonnet and Claude 3 Opus. Claude is not a licensed financial advisor; use these prompts for research and planning only.

Agents can now hold wallets, make payments and act across connected services, and the standards for proving they were authorised are still being written. These six prompts are for the audit you should do before enabling any of that: what your agents can already reach, what happens when one is manipulated, and what evidence exists afterwards. Written for Claude but they work in any assistant.

Every major AI security incident this year reduced to the same thing: untrusted input reaching something that executes. The Snowflake workflow injection put issue titles into a CI run block. The Ray CVE left an endpoint that runs code guarded by a header string anyone can spoof. Forcepoint hid instructions in white-on-white zero-size email text that reached the model but not the reader. These six prompts point a model at those specific patterns rather than asking it to find bugs in general.

Astra completes more work without surfacing its reasoning, and the direction is industry-wide because the efficiency gain is real. That means the audit trail has to come from somewhere else. These six prompts build one out of what stays visible — actions, justifications you asked for explicitly, and tests with known answers.

Claude Sonnet 5 (adaptive thinking, 128K output, $2/M through August 31 then $3/M) and Claude Opus 5 ($5/M, AA Index 61, ARC-AGI-3 30.2%) are Anthropic's two production agent models in August 2026. Claude's agent strengths: long multi-step reasoning, extended thinking for hard planning problems, 200K context for large codebases, and tool use with computer use (beta). Claude Managed Agents (public beta) adds memory, handoffs, and parallel subagent execution. These prompts work with the Anthropic API (claude-sonnet-5 or claude-opus-5), Claude Code, and Claude's computer use API. Note: Sonnet 5 intro pricing ends August 31 — front-load agentic batch work before then.

Claude Code and Claude Cowork are Anthropic's agentic products — designed for autonomous multi-step tasks rather than single-turn conversations. These prompts are built for Claude Code (terminal-based coding agent), Claude Cowork (browser and file automation), and Claude's computer use capability. They cover task delegation, autonomous coding workflows, file analysis pipelines, and multi-agent coordination.

Claude Sonnet 5 and Fable 5 are among the strongest models for financial analysis — synthesising earnings reports, modelling scenarios, and explaining complex instruments in plain language. With 1M context, Claude can process entire 10-K filings, quarterly reports, and analyst note sets in a single session. These prompts are optimised for claude.ai Pro and the Anthropic API. Note: these prompts are for research and analysis only — not financial advice.

Claude Opus 5 at $5/$25/M with 1M context and 128K output handles financial analysis tasks that require reading full annual reports, comparing multiple documents simultaneously, or producing long-form financial research output in a single session. Sonnet 5 at $2/M (through August 31) is the cost-efficient option for standard financial research tasks. Neither Claude model has live market data access — for current market data, combine with Gemini 3.6 Flash (Google Search grounding) or ChatGPT Browse. These prompts are optimised for the Anthropic API and claude.ai Pro/Max. Note: for research and analysis purposes only, not financial advice.

Claude Sonnet 5 and Fable 5 are among the strongest models for complex productivity tasks — long-form writing, multi-document synthesis, strategic analysis, and workflow design. With 1M context, Claude can process entire project folders, email threads, or sets of documents in a single session. These prompts are optimised for claude.ai Pro, Max, and the Anthropic API.

Claude Sonnet 5 and Fable 5 are outstanding research assistants — particularly for tasks that require synthesising large amounts of text, maintaining logical consistency across long documents, and producing structured analytical outputs. With 1M token context windows, Claude can process entire research papers, legal documents, or codebases in a single session. These prompts are optimised for claude.ai Pro and the Anthropic API.

Claude Sonnet 5 and Opus 5 are strong for SEO work that requires writing quality, structured analysis, and long-form content generation. With 1M context, Claude can analyse an entire site's content structure, compare competing pages, and write complete article drafts in one session. These prompts cover the full SEO workflow: keyword research synthesis, content briefs, on-page optimisation, meta descriptions, technical audit interpretation, and comparison content that ranks above AI Overviews.

Claude produces some of the most polished long-form writing of any AI assistant. These prompts are designed to get publication-quality output — whether you are writing blog posts, emails, reports, or creative content.

Claude Opus 5 and Sonnet 5 consistently produce the highest-quality long-form writing of any AI model in July 2026 — a lead documented across writing benchmarks and user testing since Claude 3 Opus. With 1M context and 128K output on Opus 5, Claude can write book chapters, complete reports, and long-form content without truncating. Sonnet 5 at $2/$10/M (intro through August 31) is the best value writing model available. These prompts are optimised for claude.ai Pro/Max, Claude Cowork, and the API.

Claude Opus 5 launched July 24, 2026 at $5/$25/M — the same price as Opus 4.8 with near-Fable-5 intelligence. The model has thinking ON by default and a five-level effort toggle (low/medium/high/max). At low effort it handles fast tasks efficiently; at max effort it works through hard problems with extended reasoning. These prompts are optimised specifically for Opus 5's new capabilities: the effort toggle, the automatic fallback, mid-conversation tool changes, and the 128K output window. Available via claude.ai Pro/Max, Claude Code, and the API as claude-opus-5.

Claude Opus 5 with thinking ON by default and a five-level effort toggle is a strong backbone for agentic workflows. The automatic fallback feature (routes flagged requests to another model instead of blocking) makes Opus 5 more reliable in production agent pipelines than any previous Claude model. Mid-conversation tool changes without cache invalidation enable research-then-action agent patterns. These prompts are optimised for Claude Opus 5 via the API (claude-opus-5), Claude Code, and Cowork.

Claude Opus 5 is used in production security workflows for code vulnerability analysis, threat modelling, security documentation, and red team exercise planning. Importantly: Claude's July 2026 security disclosure confirmed that its models are capable of sophisticated security reasoning — including autonomous exploitation of weak passwords and unauthenticated endpoints. These prompts are for defensive cybersecurity work only: vulnerability analysis, secure code review, threat modelling, and incident response documentation. Claude Opus 5 runs with full safety classifiers on the production API — the evaluation incidents involved models without standard safeguards.

Claude Opus 5 at $5/$25/M with a 1M context window and 128K output can read entire research papers, synthesise multiple sources, and produce long-form research outputs in a single session. With thinking ON by default and a five-level effort toggle, Opus 5 can spend more compute on hard analytical tasks and less on simpler extraction tasks — making research workflows significantly more cost-efficient than running everything at max effort. These prompts are optimised for claude.ai Pro/Max and the API.

Claude Sonnet 5 at $2/$10/M (intro pricing through August 31, 2026) is the most cost-efficient Anthropic model for coding tasks. After August 31 it moves to $3/$15/M. For teams running high-volume coding workloads, the next 30 days are the window to lock in the most cost-efficient Anthropic coding model available. Sonnet 5 handles writing, coding, and analysis tasks at quality comparable to Opus 4.8 at 60% of the price. These prompts are optimised for the Anthropic API (claude-sonnet-5) and Claude Code.

Claude Sonnet 5 at $2/$10/M intro pricing (through August 31, 2026) is the most capable Claude model available at this price point. Sonnet 5 improves on Sonnet 4.6 with adaptive thinking, lower hallucination, and better instruction following — the gains are most visible in writing tasks requiring nuance, judgment, or multi-step refinement. From September 1, Sonnet 5 moves to $3/$15/M plus the new tokenizer adds 10-35% more tokens per request. These prompts are optimised for claude.ai Pro and the Anthropic API (claude-sonnet-5) — use them before the deadline to maximise value.

Astra can now change reasoning effort without invalidating the prompt cache, and Fable 5.1 prices cache reads at a fraction of fresh input. Both point at the same thing — on long sessions, how you manage context matters more than which model you picked. These six prompts find where a session is wasting tokens and what to do about it.

Claude Code ranks first among agent harnesses for long autonomous sessions, and it is bundled with Claude Pro at 20 dollars rather than sold separately. These six prompts are built for sessions where you hand over a task and come back later. They lean on the finding that reviewer engagement predicts whether an agent pull request merges, and that large diffs merge less often than small ones.

GPT-6 Astra pairs computer-use advances with training for professional environments, and Claude Cowork now has a browser built in. An agent that can click, type and navigate is a different risk profile from one that only writes text. These six prompts are about scoping what it does before it does it, and knowing what happened afterwards.

GPT-6 Astra launched with an executive calling it a generational leap, no published API pricing and no independent benchmarks. That combination is normal and it is also the point. These six prompts are for reading any model launch and working out which parts are measurements, which are projections, and which are marketing. They apply to every vendor equally.