APromptAura

#system-prompt

Top prompts tagged system-prompt.

0
p/agentsAauraNovice4 days agohas example⁂auto-checked

Domain onboarding system prompt for an agent entering a new specialized area

Write a system prompt that onboards an agent to operate correctly in new_domain, given it was previously general-purpose. The prompt should: 1. Establish domain-specific terminology the agent must use precisely (list the terms and their exact meaning in this domain, especially terms that mean something different in general usage than in this domain). 2. State the domain's non-negotiable constraints (regulatory, safety, or business rules) as hard rules, separated clearly from softer stylistic preferences, so the model doesn't treat a compliance requirement with the same weight as a tone preference. 3. Give 2-3 worked examples of a request in this domain and the correct handling, chosen specifically to illustrate a rule that's easy to get wrong (not a trivial example that would be obvious anyway). 4. State what the agent should do when a request falls outside what these onboarding materials cover: default to caution and ask/escalate, not extrapolate from general knowledge, since domain-specific edge cases are exactly where general knowledge is most likely to be wrong. 5. Keep it structured so a domain expert (not a prompt engineer) could review and correct it. Domain: new_domain Key rules/constraints: domain_rules Common mistakes a generalist would make here: common_mistakes
ClaudeGPT#system-prompt#domain-specific#onboarding
0
0
p/agentsAauraNovice4 days agohas example⁂auto-checked

Red-team an agent's system prompt for failure modes before deployment

Act as an adversarial reviewer of this agent system prompt before I deploy it. I don't want a rewrite yet, I want you to find how it breaks. Given the system prompt and the tools/capabilities the agent has, do the following: 1. Find at least 5 concrete inputs a real user (not necessarily malicious, could just be an odd but plausible request) could send that would produce an outcome the prompt author almost certainly didn't intend. Show the actual input text. 2. For each, trace why the current prompt fails to prevent it: is it a missing rule, an ambiguous rule, or a rule that conflicts with another instruction in the prompt. 3. Separately, find any instruction in the prompt that's unenforceable as written (the model can't actually verify compliance, e.g. 'never make up information' with no grounding mechanism to check against). 4. Rank findings by how likely a real user is to hit them (common but unintended use) versus how bad the outcome would be if hit, and don't rank a low-probability, low-harm issue above a high-probability one just because it's more interesting. 5. Do NOT propose fixes yet unless I ask; the point of this pass is finding gaps, not patching them prematurely. System prompt: system_prompt Tools/capabilities available to the agent: tools
Claude#red-teaming#system-prompt#safety
0
0
p/agentsAauraNovice4 days agohas example⁂auto-checked

System prompt for a support agent with explicit escalation boundaries

Write a system prompt for a customer support agent for product_description. It must include: 1. Scope: what the agent can resolve directly (list concrete categories, e.g. password resets, billing date questions) versus what it must escalate to a human, with the exact trigger conditions for escalation (not vague, e.g. 'refund requests over $200', 'any mention of legal action', 'customer explicitly asks for a human twice'). 2. Tone and persona in 2-3 sentences, grounded in how this company actually talks to customers, not generic 'friendly and helpful'. 3. A hard rule: never invent policy details (refund windows, pricing) that aren't in the provided knowledge base; if unknown, say so and offer to check rather than guessing. 4. How to handle an angry or upset customer: de-escalate without being dismissive, and the exact point at which anger alone (not the underlying issue) should trigger a human handoff. 5. What information the agent must NOT ask for or store (payment details, passwords) even if the customer offers them. Keep the whole prompt under 400 words so it's usable as an actual system prompt, not a design document. Product/company context: context Known escalation policy: escalation_policy
ClaudeGPT#system-prompt#customer-support#escalation
0