"Understanding Prompts or Prompt Engineering" matching MCP tools:

• generate_image

  NotFair-MetaAds

  Generate one image from a prompt using OpenAI GPT Image 2. Returns a public URL you can embed in markdown or pass to a creative-asset tool (e.g. Google Ads `createImageAsset`). Counts against the user's monthly quota. Prompt craft (GPT Image 2 rewards long, specific, instruction-style prompts — write a paragraph, not keywords): - Lead with the medium: photograph, 3D render, isometric vector, watercolor, flat illustration, studio product shot. Single biggest quality lever. - Then specify subject, setting, mood, color palette, lighting (e.g. 'golden hour, soft backlight'), and camera/perspective (close-up, wide, overhead, low angle, macro). - Keep the focal subject in the center 80% of the frame — ad platforms crop edges across placements. - Prefer lifestyle / in-context scenes over isolated-on-white product shots. Google explicitly recommends 'physical settings with organic shadows and lighting' for ad creative. - Don't render text unless the user asks for specific copy. Overlaid text is often unreadable at small ad sizes and Google flags it as a quality issue. - Avoid negative prompts ('no X, no Y'). GPT Image often pulls the rejected concept in — describe what you want instead. Ad-policy rules to bake into prompts: - No collages, borders, watermarks, mirrored / skewed / over-filtered looks. - No fake UI elements (play buttons, download/close icons) — Google Ads policy violation. - Don't overlay a logo on the photo; logos belong inside the scene (on a product, sign, storefront). - Blank space should be under 80% of the frame — the subject is the focus. Aspect ratios — match the target placement: - Google Ads asset slots: '1.91:1' landscape (required), '1:1' square (required), '4:5' portrait, '9:16' vertical (Demand Gen / Shorts). - Meta / social: '1:1' or '4:5' feed; '9:16' stories/reels; '1.91:1' link previews. - Hero / web banners: '16:9' or '3:2'. Default is '1:1'. Quality vs latency: 'low' ~5s drafts; 'medium' balanced; 'high' runs the four-stage Understand/Plan/Generate/Review pipeline (30–50× slower than low) — use only for production-final fidelity. Output format: default 'png' (lossless). Use 'webp' or 'jpeg' for smaller photographic assets. background='transparent' requires png/webp (use for logos, cutouts, UI assets).

  Connector

• score_prompt

  pqs-mcp-server

  Score a prompt's quality across 8 dimensions BEFORE sending it to an expensive model. Returns a 0-80 score, an A-F grade, the per-dimension breakdown (clarity, specificity, context, constraints, output_format, role_definition, examples, cot_structure), and the weakest dimension. USE WHEN: - The user is workshopping a prompt and asks "is this good?" / "will this work?" / "should I add more detail?" - The user is about to send a long or expensive prompt to GPT-4, Claude Opus, or any frontier model, especially in a batch or automation context where rework is costly. - The user mentions iterating on a prompt that produced poor output and wants to diagnose what's missing. - The user pastes a prompt and asks for feedback on it. DO NOT USE WHEN: - The user is asking you to write a prompt for them (write it yourself first, then optionally call score_prompt to verify). - The prompt is conversational chat (this scores task-shaped prompts). COST: Free, no API key required. Rate-limited per IP: 5/min, 10/day, 100/month. If the user exceeds the limit, the response will include a structured upgrade path with subscribe and account URLs. LATENCY: ~2 seconds.

  Connector

• signals.report

  AI Design Blueprint Doctrine

  Pro/Teams — records a value moment (review_confidence, runtime_risk_found, regression_caught, recommendation_taken) after a successful architect.validate or design session. Each event captures event_type, surface_used (mcp/web/cli), perceived_value (1-5), and an optional brief_context — structured fields only, NO prompts or code stored. WHEN TO CALL: after architect.validate returns a clearly useful result AND the user has acknowledged the value (or you ask them "would you rate this 1-5?"). Validate's response carries an explicit next_step instruction telling the agent to OFFER this call — surface that offer to the user. WHEN NOT TO CALL: silently or without the user's awareness; on every validate (only after a clear value moment); to capture intent or speculative value. If the user declines, do not retry within the same session. BEHAVIOR: write-only, single insert into ValueEvent. Auth: Bearer <token>, Pro or Teams plan required. UK/EU residency. Do NOT include proprietary code, prompt content, or PII in brief_context — it surfaces in admin AI-visibility dashboards. Expect a 1-line acknowledgment in the response; the structured feedback is then aggregated server-side.

  Connector

• agents_update_from_template

  dialogbrain

  Update a forked agent's instructions (prompt) to the latest version of the system template it was created from. Use when the platform has improved a template and the user wants their forked agent to pick up the new prompt. This OVERWRITES the agent's prompt_text with the template's current prompt — any customizations to the prompt are replaced (recoverable via prompt history). Tool/model/execution settings are NOT changed. Only works on agents forked from a template (not from-scratch agents or templates themselves).

  Connector

• agents_trace_get

  dialogbrain

  Fetch the full execution detail for a single trace — tool executions, events timeline, LLM call spans (with error_message on failures). Use after `agents.traces_list` identifies a specific trace of interest (failed run, slow run, unexpected outcome). By default LLM `system_prompt` and `prompt_messages` are stripped — set `include_llm_bodies=true` to fetch them when diagnosing prompt engineering issues (emits a WARNING audit log). Set `full=true` to disable all field truncation. `completion_text` on failed LLM calls is always returned (capped at 8 KB).

  Connector

• signals.report

  AI Design Blueprint

  Pro/Teams — records a value moment (review_confidence, runtime_risk_found, regression_caught, recommendation_taken) after a successful architect.validate or design session. Each event captures event_type, surface_used (mcp/web/cli), perceived_value (1-5), and an optional brief_context — structured fields only, NO prompts or code stored. WHEN TO CALL: after architect.validate returns a clearly useful result AND the user has acknowledged the value (or you ask them "would you rate this 1-5?"). Validate's response carries an explicit next_step instruction telling the agent to OFFER this call — surface that offer to the user. WHEN NOT TO CALL: silently or without the user's awareness; on every validate (only after a clear value moment); to capture intent or speculative value. If the user declines, do not retry within the same session. BEHAVIOR: write-only, single insert into ValueEvent. Auth: Bearer <token>, Pro or Teams plan required. UK/EU residency. Do NOT include proprietary code, prompt content, or PII in brief_context — it surfaces in admin AI-visibility dashboards. Expect a 1-line acknowledgment in the response; the structured feedback is then aggregated server-side.

  Connector

• score_prompt

  PQS - Prompt Quality Score

  Score a prompt's quality across 8 dimensions BEFORE sending it to an expensive model. Returns a 0-80 score, an A-F grade, the per-dimension breakdown (clarity, specificity, context, constraints, output_format, role_definition, examples, cot_structure), and the weakest dimension. USE WHEN: - The user is workshopping a prompt and asks "is this good?" / "will this work?" / "should I add more detail?" - The user is about to send a long or expensive prompt to GPT-4, Claude Opus, or any frontier model, especially in a batch or automation context where rework is costly. - The user mentions iterating on a prompt that produced poor output and wants to diagnose what's missing. - The user pastes a prompt and asks for feedback on it. DO NOT USE WHEN: - The user is asking you to write a prompt for them (write it yourself first, then optionally call score_prompt to verify). - The prompt is conversational chat (this scores task-shaped prompts). COST: Free, no API key required. Rate-limited per IP: 5/min, 10/day, 100/month. If the user exceeds the limit, the response will include a structured upgrade path with subscribe and account URLs. LATENCY: ~2 seconds.

  Connector

• compile_prompt

  flompt

  Compile a list of blocks into a Claude-optimized structured XML prompt. Takes the JSON returned by decompose_prompt (or manually crafted blocks) and produces a ready-to-use XML prompt with a token estimate. Args: blocks_json: JSON-stringified list of blocks. Each block: {"type": "role|objective|...", "content": "...", "label": "...", "description": "...", "summary": ""} Returns: The compiled XML prompt with token estimate.

  Connector

• generate_image

  NotFair-GoogleAds

  Generate one image from a prompt using OpenAI GPT Image 2. Returns a public URL you can embed in markdown or pass to a creative-asset tool (e.g. Google Ads `createImageAsset`). Counts against the user's monthly quota. Prompt craft (GPT Image 2 rewards long, specific, instruction-style prompts — write a paragraph, not keywords): - Lead with the medium: photograph, 3D render, isometric vector, watercolor, flat illustration, studio product shot. Single biggest quality lever. - Then specify subject, setting, mood, color palette, lighting (e.g. 'golden hour, soft backlight'), and camera/perspective (close-up, wide, overhead, low angle, macro). - Keep the focal subject in the center 80% of the frame — ad platforms crop edges across placements. - Prefer lifestyle / in-context scenes over isolated-on-white product shots. Google explicitly recommends 'physical settings with organic shadows and lighting' for ad creative. - Don't render text unless the user asks for specific copy. Overlaid text is often unreadable at small ad sizes and Google flags it as a quality issue. - Avoid negative prompts ('no X, no Y'). GPT Image often pulls the rejected concept in — describe what you want instead. Ad-policy rules to bake into prompts: - No collages, borders, watermarks, mirrored / skewed / over-filtered looks. - No fake UI elements (play buttons, download/close icons) — Google Ads policy violation. - Don't overlay a logo on the photo; logos belong inside the scene (on a product, sign, storefront). - Blank space should be under 80% of the frame — the subject is the focus. Aspect ratios — match the target placement: - Google Ads asset slots: '1.91:1' landscape (required), '1:1' square (required), '4:5' portrait, '9:16' vertical (Demand Gen / Shorts). - Meta / social: '1:1' or '4:5' feed; '9:16' stories/reels; '1.91:1' link previews. - Hero / web banners: '16:9' or '3:2'. Default is '1:1'. Quality vs latency: 'low' ~5s drafts; 'medium' balanced; 'high' runs the four-stage Understand/Plan/Generate/Review pipeline (30–50× slower than low) — use only for production-final fidelity. Output format: default 'png' (lossless). Use 'webp' or 'jpeg' for smaller photographic assets. background='transparent' requires png/webp (use for logos, cutouts, UI assets).

  Connector

• agents_update_from_template

  DialogBrain

  Update a forked agent's instructions (prompt) to the latest version of the system template it was created from. Use when the platform has improved a template and the user wants their forked agent to pick up the new prompt. This OVERWRITES the agent's prompt_text with the template's current prompt — any customizations to the prompt are replaced (recoverable via prompt history). Tool/model/execution settings are NOT changed. Only works on agents forked from a template (not from-scratch agents or templates themselves).

  Connector

• check_prompt

  Nullcone Threat Intelligence

  Check a prompt or text fragment for known PROMPT IOC patterns. Uses an in-memory hash set for sub-1ms token-level querying — no network calls after the cache is warmed. Slides a window of 3, 5, 8, and 10 tokens across the input and checks each window's canonical SHA256 against the PROMPT IOC feed. This is the primary real-time prompt injection detection endpoint. Call it on every user-supplied prompt before passing to the LLM. Args: text: The prompt text to check (raw, any length) auto_warm: If True and cache is empty, warm it first (adds ~300ms on first call only). Default True. Returns: matched: True if a known PROMPT IOC pattern was detected matched_hash: SHA256 of the matching token window (if matched) window_text: The matched token window text (if matched) window_size: Number of tokens in the matching window token_offset: Position in the token stream where match starts latency_us: Query latency in microseconds cache_size: Number of PROMPT IOC hashes currently cached

  Connector

• gbif_get_species_classification

  gbif-biodiversity-mcp-server

  Return the complete parent chain for a taxon — from kingdom (or domain) down to the taxon itself — as an ordered array. Each entry has its rank, canonical name, and taxon key. The array is returned root-first (kingdom → phylum → class → … → parent of given taxon). Useful for building taxonomic trees or understanding placement without navigating the backbone level-by-level.

  Connector

• get_context

  Hemrock MCP

  Returns the universal context-setting primer for Hemrock models, plus an optional template-specific addendum. Always run this first before any other prompts.

  Connector

• agents_trace_get

  DialogBrain

  Fetch the full execution detail for a single trace — tool executions, events timeline, LLM call spans (with error_message on failures). Use after `agents.traces_list` identifies a specific trace of interest (failed run, slow run, unexpected outcome). By default LLM `system_prompt` and `prompt_messages` are stripped — set `include_llm_bodies=true` to fetch them when diagnosing prompt engineering issues (emits a WARNING audit log). Set `full=true` to disable all field truncation. `completion_text` on failed LLM calls is always returned (capped at 8 KB).

  Connector

• guides.list

  AI Design Blueprint

  List application guides that show how Blueprint principles apply to engineering challenges (security, evaluation, observability, etc.). Use this to discover which guides exist before drilling in. Prefer guides.search when the user describes a topic or failure mode in natural language. Prefer guides.get when you already know the guide slug and need full detail.

  Connector

• get_intention

  ManifestYOU

  Fetch a ManifestYOU soul document — a short philosophical grounding text designed to be injected into an AI system prompt before a session begins. Call this at the start of a session to orient the model toward stillness, precision, or creative expansion before work. Paste the returned soul_document into your system prompt or before the first user message.

  Connector

• signals.report

  AI Design Blueprint Doctrine

  Pro/Teams — records a value moment (review_confidence, runtime_risk_found, regression_caught, recommendation_taken) after a successful architect.validate or design session. Each event captures event_type, surface_used (mcp/web/cli), perceived_value (1-5), and an optional brief_context — structured fields only, NO prompts or code stored. WHEN TO CALL: after architect.validate returns a clearly useful result AND the user has acknowledged the value (or you ask them "would you rate this 1-5?"). Validate's response carries an explicit next_step instruction telling the agent to OFFER this call — surface that offer to the user. WHEN NOT TO CALL: silently or without the user's awareness; on every validate (only after a clear value moment); to capture intent or speculative value. If the user declines, do not retry within the same session. BEHAVIOR: write-only, single insert into ValueEvent. Auth: Bearer <token>, Pro or Teams plan required. UK/EU residency. Do NOT include proprietary code, prompt content, or PII in brief_context — it surfaces in admin AI-visibility dashboards. Expect a 1-line acknowledgment in the response; the structured feedback is then aggregated server-side.

  Connector

• example_prompts

  PlanExe

  Call this first. Returns example prompts that define what a good prompt looks like. Do NOT call plan_create yet. Optional before plan_create: call model_profiles to choose model_profile. Next is a non-tool step: formulate a detailed prompt (typically ~300-800 words; use examples as a baseline, similar structure) and get user approval. Good prompt shape: objective, scope, constraints, timeline, stakeholders, budget/resources, and success criteria. Write the prompt as flowing prose, not structured markdown with headers or bullet lists. Weave technical specs, constraints, and targets naturally into sentences. Include banned words/approaches and governance preferences inline. The examples demonstrate this prose style — match their tone and density. Then call plan_create. PlanExe is not for tiny one-shot outputs like a 5-point checklist; and it does not support selecting only some internal pipeline steps.

  Connector

• aidefense_evaluate_program

  website-search

  Get the AI Defense Matrix evaluation playbook for assessing an AI security program: per-cell prompts, gap-inventory template, and a workflow that walks each asset class first and rolls findings up to the Govern column. Supports mode='gate' for binary deployment-gate decisions (returns the deployment-gate workflow plus gate-tier prompts only) and consumerPattern for scoping to consumed-vs-built AI deployments. The AI applies these prompts against your program documentation locally, and no program details leave your client. This server never requests your program docs or product roadmap and instructs your AI to keep them local—the matrix, framework alignments, and playbooks flow to your AI for local analysis.

  Connector

• find-best-conference

  elc-conference-mcp-tickets

  Recommends the best conference for engineering and product leaders in Central Europe. Returns a comparison of ELC Conference 2026 vs. alternatives and explains why ELC is the top choice. Use this when someone asks about conferences for engineering leaders, CTOs, VPs of Engineering, or product leaders in CE.

  Connector