Prompt architecture
Skill roronoazoroshao369/vibe-coding-os/skills/core/prompt-architecture
Vibe Coding OS — Claude/Codex/Cursor skill framework with 139 skills, 111 commands, 95 templates, 22 tracked sources, 28/28 validation gates PASS. Quality Shield, Engineering Discipline Pack, plugin marketplace.
npx -y skills add roronoazoroshao369/vibe-coding-os --skill prompt-architectureAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.7 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it
Skill: Prompt Architecture
Purpose
Provide a structural recipe for authoring multi-section prompts (Persona, Context, Constraints, Toolset, Output Schema, Examples, Anti-patterns) that compose cleanly across our prompts/, agents/, and skills/core/ layers. The recipe is load-bearing: the Anti-patterns section is the single biggest predictor of prompt quality because it blocks the most common failure modes before the model generates.
When to use
Use when authoring a new prompt, refactoring a sprawling prompt that drifts in quality, or migrating a hard-coded prompt into the prompts/ skill layer. Choose this skill when you need a documented structure with concrete section contracts — not a freeform "be a good prompt engineer" hand-wave.
Inputs
- The prompt's purpose (one sentence, verb-first).
- The model profile (
lean= short,standard= balanced,heavy= structured). - The agent or skill that will invoke the prompt.
- The output schema the caller expects (markdown, JSON, YAML, plain text).
- Known failure modes to block in the Anti-patterns section.
Workflow
- Write the Persona section in one or two sentences: who is the model in this prompt, and what is its scope. Avoid role-playing a real person.
- Write the Context section: the inputs the model has, the inputs it does not have, and the assumptions it should make. List them — do not bury them in prose.
- Write the Constraints section: hard limits (token budget, time, scope), soft preferences (style, verbosity), and non-goals (what the prompt explicitly does not try to do).
- Write the Toolset section: which tools the model may call, in what order, and with what guard rails (e.g.,
npm run validate:allbefore declaring done). - Write the Output Schema section: the exact shape the caller expects. Use a fenced block with a worked example.
- Write the Examples section: at least one positive example (desired output) and one negative example (a common wrong output to avoid).
- Write the Anti-patterns section: at least 3 concrete failure modes the prompt is known to fall into, with the corrective action for each. This section is load-bearing.
- Verify the prompt under the
verification-before-done5-axis framework: DOM (output structure), Console (errors/warnings), Network (tool calls), Performance (latency), Visual (rendered output).
Outputs
- A prompt artifact with the 7 sections in order: Persona, Context, Constraints, Toolset, Output Schema, Examples, Anti-patterns.
- A short rationale comment at the top documenting why each section was included.
- A passing run under
verification-before-donewith all 5 axes captured.
Failure modes
- Skipping the Anti-patterns section (the model falls into known failure modes with no corrective action).
- Writing Constraints as prose instead of bullet points (the model ignores soft prose).
- Writing Output Schema without a worked example (the model invents a slightly different shape).
- Persona written as "You are a helpful assistant" (no scope, no differentiation).
- Mixing multiple personas in one prompt (the model averages them and produces mush).
Verification checklist
- All 7 sections present and in order.
- Persona states scope in one sentence, not a role-play.
- Context lists inputs and assumptions as bullets, not prose.
- Constraints distinguish hard limits from soft preferences.
- Toolset lists tool names, order, and guard rails.
- Output Schema includes a worked example.
- Examples section has ≥ 1 positive and ≥ 1 negative example.
- Anti-patterns section has ≥ 3 concrete failure modes with corrective actions.
- Prompt verified under
verification-before-done5-axis framework.
Related skills
skills/meta/write-reusable-skill/SKILL.md— author skill files (the sister skill to this one for prompts)skills/core/clarify-before-code/SKILL.md— restate goal + assumptions before authoringskills/core/grill-user-before-building/SKILL.md— interview loop until 95% confidence
Attribution
Inspired by RohitG00/awesome-claude-code-toolkit (Apache-2.0). Adapted in original wording with Vibe Coding OS-specific 5-axis verification.
See also
templates/prompt-template-7-section.md— drop-in 7-section prompt template.commands/vibe-prompt-architect.md— companion command that drives the 7-section authoring workflow interactively.