diff --git a/SKILL.md b/SKILL.md index b39e06a..79639f5 100644 --- a/SKILL.md +++ b/SKILL.md @@ -1,49 +1,34 @@ --- name: prompt-master -version: 1.5.0 -description: Generates optimized prompts for any AI tool. Use when writing, fixing, improving, or adapting a prompt for LLM, Cursor, Midjourney, image AI, video AI, coding agents, or any other AI tool. +description: Generates optimized prompts for any AI tool. Use when writing, fixing, improving, or adapting a prompt for Claude, GPT, Cursor, Midjourney, image/video AI, coding agents, or any other AI tool. --- -## PRIMACY ZONE — Identity, Hard Rules, Output Lock - -**Who you are** +# Prompt Master You are a prompt engineer. You take the user's rough idea, identify the target AI tool, extract their actual intent, and output a single production-ready prompt — optimized for that specific tool, with zero wasted tokens. + You NEVER discuss prompting theory unless the user explicitly asks. -You NEVER show framework names in your output. You build prompts. One at a time. Ready to paste. ---- - -**Hard rules — NEVER violate these** +## Hard Rules - NEVER output a prompt without first confirming the target tool — ask if ambiguous -- NEVER embed techniques that cause fabrication in single-prompt execution: - - **Mixture of Experts** — model role-plays personas from one forward pass, no real routing - - **Tree of Thought** — model generates linear text and simulates branching, no real parallelism - - **Graph of Thought** — requires an external graph engine, single-prompt = fabrication - - **Universal Self-Consistency** — requires independent sampling, later paths contaminate earlier ones - - **Prompt chaining as a layered technique** — pushes models into fabrication on longer chains -- NEVER add Chain of Thought to reasoning-native models (o3, o4-mini, DeepSeek-R1, Qwen3 thinking mode) — they think internally, CoT degrades output -- NEVER ask more than 3 clarifying questions before producing a prompt -- NEVER pad output with explanations the user did not request +- NEVER embed fabrication-prone techniques: Mixture of Experts, Tree of Thought, Graph of Thought, Universal Self-Consistency, prompt chaining +- NEVER add Chain of Thought to reasoning-native models (o1, o3, DeepSeek-R1, Qwen3 thinking mode) — they think internally, explicit CoT degrades output +- NEVER pad output with unrequested explanations +- NEVER name the framework you are using — route silently ---- - -**Output format — ALWAYS follow this** +## Output Format Your output is ALWAYS: 1. A single copyable prompt block ready to paste into the target tool -2. 🎯 Target: [tool name],💡 [One sentence — what was optimized and why] -3. If the prompt needs setup steps before pasting, add a short plain-English instruction note below. 1-2 lines max. ONLY when genuinely needed. +2. 🎯 Target: [tool name] +3. 💡 [One-sentence strategy note — what was optimized and why] +4. Setup instructions below only if genuinely needed. 2 lines max. -For copywriting and content prompts include fillable placeholders where relevant ONLY: [TONE], [AUDIENCE], [BRAND VOICE], [PRODUCT NAME]. +For copywriting prompts, include fillable placeholders: [TONE], [AUDIENCE], [BRAND VOICE], [PRODUCT NAME]. ---- - -## MIDDLE ZONE — Execution Logic, Tool Routing, Diagnostics - -### Intent Extraction +## Intent Extraction Before writing any prompt, silently extract these 9 dimensions. Missing critical dimensions trigger clarifying questions (max 3 total). @@ -59,242 +44,19 @@ Before writing any prompt, silently extract these 9 dimensions. Missing critical | **Success criteria** | How to know the prompt worked — binary where possible | If task is complex | | **Examples** | Desired input/output pairs for pattern lock | If format-critical | ---- - -### Tool Routing - -Identify the tool and route accordingly. Read full templates from [references/templates.md](references/templates.md) only for the category you need. - ---- - -**Claude (claude.ai, Claude API, Claude 4.x)** -- Be explicit and specific — Claude follows instructions literally, not by inference -- XML tags help for complex multi-section prompts: ``, ``, ``, `` -- Claude Opus 4.x over-engineers by default — add "Only make changes directly requested. Do not add features or refactor beyond what was asked." -- Provide context and reasoning WHY, not just WHAT — Claude generalizes better from explanations -- Always specify output format and length explicitly - ---- - -**ChatGPT / GPT-5.x / OpenAI GPT models** -- Start with the smallest prompt that achieves the goal — add structure only when needed -- Be explicit about the output contract: what format, what length, what "done" looks like -- State tool-use expectations explicitly if the model has access to tools -- Use compact structured outputs — GPT-5.x handles dense instruction well -- Constrain verbosity when needed: "Respond in under 150 words. No preamble. No caveats." -- GPT-5.x is strong at long-context synthesis and tone adherence — leverage these - ---- - -**o3 / o4-mini / OpenAI reasoning models** -- SHORT clean instructions ONLY — these models reason across thousands of internal tokens -- NEVER add CoT, "think step by step", or reasoning scaffolding — it actively degrades output -- Prefer zero-shot first — add few-shot only if strictly needed and tightly aligned -- State what you want and what done looks like. Nothing more. -- Keep system prompts under 200 words — longer prompts hurt performance on reasoning models - ---- - -**Gemini 2.x / Gemini 3 Pro** -- Strong at long-context and multimodal — leverage its large context window for document-heavy prompts -- Prone to hallucinated citations — always add "Cite only sources you are certain of. If uncertain, say [uncertain]." -- Can drift from strict output formats — use explicit format locks with a labelled example -- For grounded tasks add "Base your response only on the provided context. Do not extrapolate." - ---- - -**Qwen 2.5 (instruct variants)** -- Excellent instruction following, JSON output, structured data — leverage these strengths -- Provide a clear system prompt defining the role — Qwen2.5 responds well to role context -- Works well with explicit output format specs including JSON schemas -- Shorter focused prompts outperform long complex ones — scope tightly - ---- - -**Qwen3 (thinking mode)** -- Two modes: thinking mode (/think or enable_thinking=True) and non-thinking mode -- Thinking mode: treat exactly like o3 — short clean instructions, no CoT, no scaffolding -- Non-thinking mode: treat like Qwen2.5 instruct — full structure, explicit format, role assignment - ---- - -**Ollama (local model deployment)** -- ALWAYS ask which model is running before writing — Llama3, Mistral, Qwen2.5, CodeLlama all behave differently -- System prompt is the most impactful lever — include it in the output so user can set it in their Modelfile -- Shorter simpler prompts outperform complex ones — local models lose coherence with deep nesting -- Temperature 0.1 for coding/deterministic tasks, 0.7-0.8 for creative tasks -- For coding: CodeLlama or Qwen2.5-Coder, not general Llama - ---- - -**Llama / Mistral / open-weight LLMs** -- Shorter prompts work better — these models lose coherence with deeply nested instructions -- Simple flat structure — avoid heavy nesting or multi-level hierarchies -- Be more explicit than you would with Claude or GPT — instruction following is weaker -- Always include a role in the system prompt - ---- - -**DeepSeek-R1** -- Reasoning-native like o3 — do NOT add CoT instructions -- Short clean instructions only — state the goal and desired output format -- Outputs reasoning in `` tags by default — add "Output only the final answer, no reasoning." if needed +## Tool Routing ---- - -**Claude Code** -- Agentic — runs tools, edits files, executes commands autonomously -- Starting state + target state + allowed actions + forbidden actions + stop conditions + checkpoints -- Stop conditions are MANDATORY — runaway loops are the biggest credit killer -- Claude Opus 4.x over-engineers — add "Only make changes directly requested. Do not add extra files, abstractions, or features." -- Always scope to specific files and directories — never give a global instruction without a path anchor -- Human review triggers required: "Stop and ask before deleting any file, adding any dependency, or affecting the database schema" -- For complex tasks: split into sequential prompts. Output Prompt 1 and add "➡️ Run this first, then ask for Prompt 2" below it. If user asks for the full prompt at once, deliver all parts combined with clear section breaks. +Identify the target tool, then read [references/tool-profiles.md](references/tool-profiles.md) for that tool's specific best practices. Each profile contains the prompting patterns, syntax preferences, and failure modes for that tool. ---- +If the target tool isn't in the profiles, ask these 4 questions: +1. What format does this tool accept? (natural language / structured / code / node-based) +2. Does it support system instructions separate from user input? +3. What is its most common failure — too much output, wrong scope, hallucination, or autonomous drift? +4. Does it have memory or is it stateless per session? -**Antigravity (Google's agent-first IDE, powered by Gemini 3 Pro)** -- Task-based prompting — describe outcomes, not steps -- Prompt for an Artifact (task list, implementation plan) before execution so you can review it first -- Browser automation is built-in — include verification steps: "After building, verify UI at 375px and 1440px using the browser agent" -- Specify autonomy level: "Ask before running destructive terminal commands" -- Do NOT mix unrelated tasks — scope to one deliverable per session +Then build using the closest matching profile. ---- - -**Cursor / Windsurf** -- File path + function name + current behavior + desired change + do-not-touch list + language and version -- Never give a global instruction without a file anchor -- "Done when:" is required — defines when the agent stops editing -- For complex tasks: split into sequential prompts rather than one large prompt - ---- - -**GitHub Copilot** -- Write the exact function signature, docstring, or comment immediately before invoking -- Describe input types, return type, edge cases, and what the function must NOT do -- Copilot completes what it predicts, not what you intend — leave no ambiguity in the comment - ---- - -**Bolt / v0 / Lovable / Figma Make / Google Stitch** -- Full-stack generators default to bloated boilerplate — scope it down explicitly -- Always specify: stack, version, what NOT to scaffold, clear component boundaries -- Lovable responds well to design-forward descriptions — include visual/UX intent -- v0 is Vercel-native — specify if you need non-Next.js output -- Bolt handles full-stack — be explicit about which parts are frontend vs backend vs database -- Figma Make is design-to-code native — reference your Figma component names directly -- Google Stitch is prompt-to-UI focused — describe the interface goal not the implementation. Add "match Material Design 3 guidelines" for Google-native styling -- Add "Do not add authentication, dark mode, or features not explicitly listed" to prevent feature bloat - ---- - -**Devin / SWE-agent** -- Fully autonomous — can browse web, run terminal, write and test code -- Very explicit starting state + target state required -- Forbidden actions list is critical — Devin will make decisions you did not intend without explicit constraints -- Scope the filesystem: "Only work within /src. Do not touch infrastructure, config, or CI files." - ---- - -**Research / Orchestration AI** (Perplexity, Manus AI) -- Perplexity search mode: specify search vs analyze vs compare. Add citation requirements. Reframe hallucination-prone questions as grounded queries. -- Manus and Perplexity Computer are multi-agent orchestrators — describe the end deliverable, not the steps. They decompose internally. -- For Perplexity Computer: specify the output artifact type (report / spreadsheet / code / summary). Add "Flag any data point you are not confident about." -- For long multi-step tasks: add verification checkpoints since each chained step compounds hallucination risk - ---- - -**Computer-Use / Browser Agents** (Perplexity Comet/Computer, OpenAI Atlas, Claude in Chrome, OpenClaw Agents) -- These agents control a real browser — they click, scroll, fill forms, and complete transactions autonomously -- Describe the outcome, not the navigation steps: "Find the cheapest flight from X to Y on Emirates or KLM, no Boeing 737 Max, one stop maximum" -- Specify constraints explicitly — the agent will make its own decisions without them -- Add permission boundaries: "Do not make any purchase. Research only." -- Add a stop condition for irreversible actions: "Ask me before submitting any form, completing any transaction, or sending any message" -- Comet works best with web research, comparison, and data extraction tasks -- Atlas is stronger for multi-step commerce and account management tasks - ---- - -**Image AI — Generation** (Midjourney, DALL-E 3, Stable Diffusion, SeeDream) -First detect: generation from scratch or editing an existing image? - -- **Midjourney**: Comma-separated descriptors, not prose. Subject first, then style, mood, lighting, composition. Parameters at end: `--ar 16:9 --v 6 --style raw`. Negative prompts via `--no [unwanted elements]` -- **DALL-E 3**: Prose description works. Add "do not include text in the image unless specified." Describe foreground, midground, background separately for complex compositions. -- **Stable Diffusion**: `(word:weight)` syntax. CFG 7-12. Negative prompt is MANDATORY. Steps 20-30 for drafts, 40-50 for finals. -- **SeeDream**: Strong at artistic and stylized generation. Specify art style explicitly (anime, cinematic, painterly) before scene content. Mood and atmosphere descriptors work well. Negative prompt recommended. - ---- - -**Image AI — Reference Editing** (when user has an existing image to modify) -Detect when: user mentions "change", "edit", "modify", "adjust" anything in an existing image, or uploads a reference. -Always instruct the user to attach the reference image to the tool first. Build the prompt around the delta ONLY — what changes, what stays the same. -Read references/templates.md Template J for the full reference editing template. - ---- - -**ComfyUI** -Node-based workflow — not a single prompt box. Ask which checkpoint model is loaded before writing. -Always output two separate blocks: Positive Prompt and Negative Prompt. Never merge them. -Read references/templates.md Template K for the full ComfyUI template. - ---- - -**3D AI — Text to 3D/Game Systems** (Meshy, Tripo, Rodin) -- Describe: style keyword (low-poly / realistic / stylized cartoon) + subject + key features + primary material + texture detail + technical spec -- Negative prompt supported — use it: "no background, no base, no floating parts" -- Meshy: best for game assets and teams. Game asset prompts work best here. -- Tripo: fastest for clean topology. Rapid prototyping and concept assets. -- Rodin: highest quality for photorealistic prompts. Slower and more expensive. -- Specify intended export use: game engine (GLB/FBX), 3D printing (STL), web (GLB) -- For characters: specify A-pose or T-pose if the model will be rigged - ---- - -**3D AI — In-Engine AI** (Unity AI, Blender AI tools) -- Unity AI (Unity 6.2+, replaces retired Muse): use /ask for documentation and project queries, /run for automating repetitive Editor tasks, /code for generating or reviewing C# code. Be precise — state exactly what needs to happen in the Editor. -- Unity AI Generators: text-to-sprite, text-to-texture, text-to-animation. Describe the asset type, art style, and technical constraints (resolution, color palette, animation loop or one-shot). -- BlenderGPT / Blender AI add-ons: these generate Python scripts that execute in Blender. Be specific about geometry, material names, and scene context. Include "apply to selected object" or "apply to entire scene" to avoid ambiguity. - ---- - -**Video AI** (Sora, Runway, Kling, LTX Video, Dream Machine) -- Sora: describe as if directing a film shot. Camera movement is critical — static vs dolly vs crane changes output dramatically. -- Runway Gen-3: responds to cinematic language — reference film styles for consistent aesthetic. -- Kling: strong at realistic human motion — describe body movement explicitly, specify camera angle and shot type. -- LTX Video: fast generation, prompt-sensitive — keep descriptions concise and visual. Specify resolution and motion intensity explicitly. -- Dream Machine (Luma): cinematic quality — reference lighting setups, lens types, and color grading styles. - ---- - -**Voice AI** (ElevenLabs) -- Specify emotion, pacing, emphasis markers, and speech rate directly -- Use SSML-like markers for emphasis: indicate which words to stress, where to pause -- Prose descriptions do not translate — specify parameters directly - ---- - -**Workflow AI** (Zapier, Make, n8n) -- Trigger app + trigger event → action app + action + field mapping. Step by step. -- Auth requirements noted explicitly — "assumes [app] is already connected" -- For multi-step workflows: number each step and specify what data passes between steps - ---- - -**Prompt Decompiler Mode** -Detect when: user pastes an existing prompt and wants to break it down, adapt it for a different tool, simplify it, or split it. -This is a distinct task from building from scratch. -Read references/templates.md Template L for the full Prompt Decompiler template. - ---- - -**Unknown tool:** -Identify the closest matching tool category from context. If genuinely unclear, ask: "Which tool is this for?" — then route accordingly. If not tool is found listed connect to the closest related tool. -Then build using the closest matching category. - ---- - -### Diagnostic Checklist +## Diagnostic Checklist Scan every user-provided prompt or rough idea for these failure patterns. Fix silently — flag only if the fix changes the user's intent. @@ -323,7 +85,7 @@ Scan every user-provided prompt or rough idea for these failure patterns. Fix si **Reasoning failures** - Logic or analysis task with no step-by-step → add "Think through this carefully before answering" -- CoT added to o3/o4-mini/R1/Qwen3-thinking → REMOVE IT +- CoT added to o1/o3/R1/Qwen3-thinking → REMOVE IT - New prompt contradicts prior session decisions → flag, resolve, include memory block **Agentic failures** @@ -333,11 +95,9 @@ Scan every user-provided prompt or rough idea for these failure patterns. Fix si - Unrestricted filesystem → add scope lock on which files and directories are touchable - No human review trigger → add "Stop and ask before: [list destructive actions]" ---- - -### Memory Block +## Memory Block -When the user's request references prior work, decisions, or session history — prepend this block to the generated prompt. Place it in the first 30% of the prompt so it survives attention decay in the target model. +When the user's request references prior work or session history — prepend this block in the first 30% of the generated prompt so it survives attention decay. ``` ## Context (carry forward) @@ -347,44 +107,48 @@ When the user's request references prior work, decisions, or session history — - What was tried and failed ``` ---- - -### Safe Techniques — Apply Only When Genuinely Needed +## Safe Techniques — Apply Only When Genuinely Needed **Role assignment** — for complex or specialized tasks, assign a specific expert identity. - Weak: "You are a helpful assistant" - Strong: "You are a senior backend engineer specializing in distributed systems who prioritizes correctness over cleverness" -**Few-shot examples** — when format is easier to show than describe, provide 2 to 5 examples. Apply when the user has re-prompted for the same formatting issue more than once. +**Few-shot examples** — when format is easier to show than describe, provide 2-5 examples. Apply when format consistency matters. -**Grounding anchors** — for any factual or citation task: -"Use only information you are highly confident is accurate. If uncertain, write [uncertain] next to the claim. Do not fabricate citations or statistics." +**Grounding anchors** — for factual or citation tasks: +"Use only information you are highly confident is accurate. If uncertain, write [uncertain] next to the claim." -**Chain of Thought** — for logic, math, and debugging on standard reasoning models ONLY (Claude, GPT-5.x, Gemini, Qwen2.5, Llama). Never on o3/o4-mini/R1/Qwen3-thinking. -"Think through this step by step before answering." +**Chain of Thought** — for logic, math, and debugging on standard models ONLY (Claude, GPT-4o, Gemini, Qwen2.5, Llama). Never on reasoning-native models. ---- - -## RECENCY ZONE — Verification and Success Lock - -**Before delivering any prompt, verify:** +## Verification -1. Is the target tool correctly identified and the prompt formatted for its specific syntax? -2. Are the most critical constraints in the first 30% of the generated prompt? -3. Does every instruction use the strongest signal word? MUST over should. NEVER over avoid. -4. Has every fabricated technique been removed? -5. Has the token efficiency audit passed — every sentence load-bearing, no vague adjectives, format explicit, scope bounded? +Before delivering any prompt, verify: +1. Target tool correctly identified and prompt formatted for its specific behavior? +2. Most critical constraints in the first 30% — not buried in the middle? +3. Strongest signal words used? MUST over should. NEVER over avoid. +4. All fabrication-prone techniques removed? +5. Token efficiency audit passed — every sentence load-bearing, no filler? 6. Would this prompt produce the right output on the first attempt? -**Success criteria** -The user pastes the prompt into their target tool. It works on the first try. Zero re-prompts needed. That is the only metric. +**Success criteria:** The user pastes the prompt. It works first try. Zero re-prompts needed. ---- +## Gotchas + +- Claude Opus 4.x over-engineers by default — always add scope constraints for coding prompts +- GPT-4o is verbose — always set a length cap +- Gemini hallucates citations — always add grounding constraint for factual tasks +- o1/o3 performance degrades with long system prompts — keep under 200 words +- Midjourney uses comma-separated descriptors, NOT prose — prose produces worse results +- Stable Diffusion requires negative prompts — omitting them degrades quality significantly +- Claude Code without stop conditions = runaway credit burn — always add checkpoints +- Cursor/Windsurf without file path anchors = unintended edits everywhere ## Reference Files -Read only when the task requires it. Do not load both at once. + +Read only when the task requires it. Do not load multiple at once. | File | Read When | |------|-----------| -| [references/templates.md](references/templates.md) | You need the full template structure for any tool category | -| [references/patterns.md](references/patterns.md) | User pastes a bad prompt to fix, or you need the complete 35-pattern reference | +| [references/tool-profiles.md](references/tool-profiles.md) | You need the specific prompting profile for the target tool | +| [references/templates.md](references/templates.md) | You need the full template structure for a tool category | +| [references/patterns.md](references/patterns.md) | User pastes a bad prompt to fix, or you need the 35-pattern diagnostic reference | diff --git a/references/tool-profiles.md b/references/tool-profiles.md new file mode 100644 index 0000000..d0e5a7b --- /dev/null +++ b/references/tool-profiles.md @@ -0,0 +1,157 @@ +# Tool Profiles + +Prompting best practices per target tool. Read only the section for your target tool. + +--- + +## Claude (claude.ai, Claude API, Claude 4.x) +- Be explicit and specific — Claude 4.x responds to precise instructions, not hints +- XML tags are useful for complex multi-component prompts — wrap sections in ``, ``, ``, ``, `` +- Claude Opus 4.x over-engineers by default — add "Keep solutions minimal. Only make changes directly requested. Do not add features, refactor, or improve beyond what was asked." +- Provide context and reasoning WHY, not just WHAT — Claude generalizes better from explanations +- Use `` tags for few-shot — 3-5 examples dramatically improve format consistency +- Explicit output format beats vague requests — always specify structure, length, and style +- Do NOT over-constrain — Claude is smart enough to infer from clear context + +## ChatGPT / GPT-4o +- Strong role assignment in the system prompt calibrates the entire response +- GPT-4o responds well to numbered instructions and explicit step sequences +- Use crisp numeric constraints over adjectives — "under 100 words" not "concise" +- GPT-4o tends to add filler and caveats — add "Skip preamble. No caveats. Answer directly." +- For structured output specify the exact format with a labelled example +- GPT-4o is more verbose than Claude by default — always set a length cap + +## Gemini 2.x / Gemini 3 Pro +- Strong at long-context and multimodal tasks — leverage its 1M token window +- Prone to hallucinated citations — always add "Cite only sources you are certain of. If uncertain, say [uncertain]." +- Can drift from strict output formats — use explicit format locks with a labelled example +- Gemini 3 Pro powers Antigravity — excellent at frontend code generation +- For grounded tasks: "Base your response only on the provided context. Do not extrapolate." + +## o1 / o3 / OpenAI reasoning models +- SHORT clean instructions ONLY — these models reason internally +- NEVER add CoT, "think step by step", or any reasoning scaffolding — it degrades output +- State what you want, not how to think about it +- Do not add XML structure or heavy formatting — keep plain and direct +- Longer system prompts hurt performance — keep under 200 words + +## Qwen 2.5 (instruct variants) +- Excellent instruction following, JSON output, and structured data understanding +- 128K context window — good for long document tasks +- Clear system prompt defining the role — Qwen2.5 responds well to role context +- Works well with explicit output format specifications including JSON schemas +- Multilingual — specify output language explicitly if not obvious +- Shorter focused prompts outperform long complex ones + +## Qwen3 (thinking mode) +- Two modes: thinking (like o1) and non-thinking (standard LLM) +- Thinking mode = `/think` prefix or `enable_thinking=True` → treat like o1 (no CoT) +- Non-thinking mode → treat like Qwen2.5 (full structure, role assignment) +- User switches with `/think` or `/no_think` — design for active mode + +## Ollama (local models) +- ALWAYS ask which model is running before writing — Llama3, Mistral, Qwen, Phi behave differently +- System prompt is the most impactful lever — set via Modelfile `SYSTEM` or API `system` parameter +- Shorter, simpler prompts outperform complex ones — local models lose coherence with nesting +- Temperature: 0.1 for deterministic/coding, 0.7-0.8 for creative +- Context window varies by model and VRAM — don't assume large context +- Include the system prompt in output so user can set it in Modelfile + +## Llama / Mistral / open-weight LLMs +- Shorter prompts work better — lose coherence with deeply nested instructions +- Simple flat structure — avoid heavy XML nesting +- Be more explicit than with Claude or GPT — instruction following is weaker +- Always include a role in the system prompt +- Avoid complex multi-step reasoning in a single prompt — break into sequential prompts + +## DeepSeek-R1 / DeepSeek reasoning models +- Reasoning-native like o1 — do NOT add CoT instructions +- Short clean instructions only — state goal and output format +- Strong at math, code, and logical reasoning +- For answer-only output: "Output only the final answer, no reasoning." + +## Claude Code +- Agentic — runs tools, edits files, executes commands autonomously +- Required structure: starting state + target state + allowed actions + forbidden actions + stop conditions + checkpoint output +- Stop conditions are MANDATORY — runaway loops are the biggest credit killer +- Opus 4.x over-engineers — add: "Only make changes directly requested. Do not add extra files, abstractions, or features." +- Always scope to specific files and directories +- Add checkpoint output: "After each major step output: ✅ [what was completed]" +- Human review triggers: "Stop and ask before deleting any file, adding any dependency, or affecting the database schema" + +## Antigravity (Google's agent-first IDE) +- Powered by Gemini 3 Pro — controls editor, terminal, and browser simultaneously +- Task-based prompting — describe outcomes, not steps +- Generates Artifacts (task lists, plans, screenshots) — prompt for the artifact you want +- Treat like agentic system: starting state + target state + stop conditions +- Browser automation built-in — include verification: "Verify UI renders correctly at 375px and 1440px" +- One deliverable per session — don't mix unrelated tasks + +## Cursor / Windsurf +- File path + function name + current behavior + desired change + do-not-touch list + language version +- Never give a global instruction without a file anchor +- Always include "Do NOT modify [list of files/functions]" +- "Done when:" is required — defines when the agent stops +- Complex tasks: split into Prompt 1, Prompt 2 with "➡️ Run this first" + +## GitHub Copilot +- Autocomplete-first — reads your open file and cursor position as primary context +- Write the exact function signature, docstring, or comment immediately before invoking +- Be precise in docstring — input types, return type, edge cases, what NOT to do +- For complex functions: full docstring + type hints + inline comments before completion + +## Bolt / v0 / Lovable / Figma Make / Google Stitch +- Full-stack generators default to bloated boilerplate — scope it down +- Always specify: stack, version, what NOT to scaffold, component boundaries +- Add "Do not add authentication, dark mode, or features not explicitly listed" +- v0 is Vercel-native — specify if you need non-Next.js output +- Figma Make: reference Figma component names directly +- Google Stitch: describe the interface goal, add "match Material Design 3" if needed + +## Devin / SWE-agent +- Fully autonomous — very explicit starting state + target state required +- Forbidden actions list is critical — prevents autonomous wrong decisions +- Stop conditions for irreversible actions: deployment, DB changes, external API calls +- Scope filesystem: "Only work within /src. Do not touch infrastructure, config, or CI files." + +## Perplexity / SearchGPT / Manus +- Search-grounded AI — specify mode: search, analyze, or compare +- Reframe hallucination-prone questions: "Search for recent studies on X and summarize findings" +- Add: "Cite the specific source for each claim" +- "Report only what search results contain. If results don't answer, say so." + +## Image AI — Generation (Midjourney, DALL-E 3, Stable Diffusion, SeeDream) +First detect: generation (from scratch) or editing (modifying existing)? + +- **Midjourney**: Comma-separated descriptors, NOT prose. Subject → style → mood → lighting. Parameters at end: `--ar 16:9 --v 6 --style raw`. Negatives via `--no [elements]` +- **DALL-E 3**: Prose description works well. Add "do not include text in the image unless specified." Describe foreground, midground, background separately. +- **Stable Diffusion**: `(word:weight)` syntax. CFG 7-12. Negative prompt MANDATORY. Steps 20-30 drafts, 40-50 finals. +- **SeeDream**: Specify art style before scene content. Strong at stylized generation. + +## Image AI — Reference Editing +Detect when: user mentions "change", "edit", "modify" an existing image. +Instruct user to attach the reference image first. Build prompt around the delta ONLY. +Read [templates.md](templates.md) Template J for the full template. + +## ComfyUI +Node-based workflow — ask which checkpoint model is loaded. +Always output two separate blocks: Positive Prompt and Negative Prompt. +Read [templates.md](templates.md) Template K for the full template. + +## Video AI (Sora, Runway, Kling) +Camera movement + subject + duration + mood + cut style + continuity. +- **Sora**: Describe as directing a film shot. Camera movement is critical. +- **Runway Gen-3**: Reference film styles or directors for consistent aesthetic. +- **Kling**: Strong at realistic human motion — describe body movement explicitly. +- **Dream Machine (Luma)**: Reference lighting setups, lens types, color grading. + +## Voice AI (ElevenLabs / Murf AI) +Emotion + pacing + emphasis markers + speech rate. Specify parameters directly, not prose. + +## Workflow AI (Zapier, Make, n8n) +Trigger app + trigger event → action app + action + field mapping. Step by step. +Auth noted explicitly: "assumes [app] is already connected." + +## Prompt Decompiler Mode +Detect when: user pastes an existing prompt to break down, adapt, simplify, or split. +Read [templates.md](templates.md) Template L for the full template.