Llama prompts are written for Meta's open Llama models, popular for self-hosted and privacy-conscious workflows. Prompt Dock human-reviews every Llama prompt, so the instructions you copy have been human-reviewed rather than guessed. Browse by category — coding, writing, analysis, agents — reveal the full prompt, and adapt it to whichever Llama variant or fine-tune you run. Free and premium Llama prompts are both available, each labelled with its recommended model so you can match the prompt to your deployment.
14 prompts
Give a talk title and get three different opening hooks built to grab a live audience in the first thirty seconds — a story, a question, a surprising fact. Start strong instead of with throat-clearing, and earn the attention the rest of your talk needs.
Give it a slow query and your schema, and it returns a faster rewrite with a clear explanation of why the original was slow and exactly which index to add. Built for the moment a dashboard times out and you need the fix, not a lecture.
Paste study notes and get fifteen spaced-repetition flashcards built for active recall — each a tight question and a precise answer, not a copy-paste of the text. The fastest path from reading something once to actually remembering it later.
Name your product and get a twelve-question interview guide built to capture a real customer story — the before, the turning point, the results, and the quotable moments. Walk into the call ready to get a case study that actually persuades.
Name a role and get an inclusive, specific job description — mission, real responsibilities, must-haves versus nice-to-haves — written to attract strong, diverse candidates. Clear enough to draw the right people and screen out the wrong fit early.
Give one observation and get five distinct, testable hypotheses that could explain it — each paired with a metric and a way to test it. Turns a curious what-just-happened into a real experiment plan instead of a single hunch you fixate on.
Paste student work and get five constructive feedback comments the student can actually act on — specific, kind, and tied to what's on the page. Saves you writing the same notes from scratch and keeps the feedback genuinely useful.
Name a role and get a structured, bias-resistant interview scorecard — five competencies that genuinely predict success, each with what to probe and how to rate it. Replaces gut-feel hiring with a fair, consistent yardstick everyone uses.
Name a habit and get a realistic four-week plan grounded in habit science — the cue, the routine, the reward, and how to recover from a missed day. A plan built for how behavior actually sticks, not a burst of motivation that fades by week two.
Paste a pile of scattered signals and get a sharp one-page trend brief that ties them into a single coherent story — what's happening, why now, and what it means for you. Synthesis, not a list, so the pattern is actually usable.
Paste a chapter or syllabus section and get a focused, exam-ready study guide that highlights what matters and how it connects — so a learner reviews efficiently instead of rereading everything. The difference between studying and just rereading.
Name a method, path, and framework, and get a clean, production-shaped endpoint with typed request and response models, input validation, sensible error handling, and a test. The boilerplate is done, so you can go straight to the real logic.
Give it your email topic and get a dozen subject lines, each paired with preview text and grouped by angle — curiosity, benefit, urgency, and more. A ready-made test menu for the part of the email that decides whether anyone opens it.
Paste an email thread and get a single concise, on-tone reply the recipient can send with barely an edit. It reads the context, matches the register, and gets to the point — clearing the message that's been sitting in your drafts for days.
Open models are more sensitive to prompt phrasing than hosted frontier models, so a reviewed, working instruction pattern saves real iteration time. Each Llama prompt here was checked by a human before publishing.
They're written for the variant named on the listing and generally adapt well — expect to tune system-prompt framing or stop tokens for your own deployment.
Free listings can be revealed by any account at no cost; premium listings are included with an Explorer or Master subscription.