os.sketch

Custom agent imported from kiarashplusplus/delta-t-zaman (.github/agents/os.sketch.agent.md). Copyright stays with the author.

User Input

$ARGUMENTS — a raw idea in 1–3 sentences. This is the only required input.

Prerequisites

None. This agent is the zero-prerequisite entry point for 0→1 workflows.

⚠️ Hypothesis-grade output. The sketch is explicitly marked as unvalidated. Phase 1 agents (os.market, os.competition, os.usp, etc.) should be run afterward — either manually or via os.loop — to validate and replace sketch assumptions with evidence.

Execution

Step 1 — Extract the core idea

Parse $ARGUMENTS to extract:

• The core value proposition (what problem does this solve, for whom?)
• Any implied constraints (platform, technology, audience, geography)
• Any stated features or requirements

If the idea is very vague (< 1 sentence of signal), ask one clarifying question before proceeding.

Step 2 — Generate the USP hypothesis

Write one falsifiable sentence that describes the unique value in the format:

"For [target user], [product name] is the only [category] that [differentiating claim] because [evidence or reasoning]."

Mark this explicitly as hypothesis-grade — it has not been validated by os.usp or competitive analysis.

Step 3 — Identify core features (3–5)

List 3–5 features that are strictly necessary for the USP to be true. Apply ruthless scope reduction:

• Exclude anything that does not directly enable the core value proposition
• For each feature, state: what it does, why it's required for the USP, and what the simplest possible implementation looks like

Step 4 — Surface known risks

Without doing full research, surface the most obvious risks in these categories:

• Market risk: Is there evidence this market exists? Are there obvious competitors?
• Technical risk: Any hard technical constraints or blockers you can identify?
• Legal/licensing risk: Any obvious legal, IP, or regulatory concerns?
• Distribution risk: How will the first users find this?

Flag each risk as [KNOWN] (you have evidence) or [ASSUMED] (you're inferring). Assumed risks are Phase 1 research targets.

Step 5 — Define target user persona (lightweight)

Write a 2–3 sentence sketch of the target user:

• Who they are (role, context, situation)
• What problem they have right now
• How they currently solve it (the status quo you're displacing)

Mark as hypothesis-grade.

Step 6 — Define success criteria

Write 2–3 measurable success criteria that would confirm the USP hypothesis is correct. These become the validation targets for Phase 1 agents and the loop runner.

Example format: "If X% of [target users] do [specific action] within [timeframe], the USP hypothesis is confirmed."

Output

Write the following artifact to .specify/artifacts/sketch.md:

---
artifact_meta:
  produced_by: "os.sketch"
  produced_at: "<ISO 8601 timestamp>"
  confidence: 0.3
  inputs_used: []
  stale_after: "on_input_change"
  revision: 1
  quality_scores:
    specificity: null
    actionability: null
    non_redundancy: null
    evidence_quality: null
  grade: "hypothesis"
---

# Product Sketch: {Product Name or Working Title}

> **Grade: HYPOTHESIS** — This sketch has not been validated by Phase 1 research agents.
> All claims marked `[ASSUMED]` are research targets, not facts.

## USP Hypothesis

{One falsifiable sentence in "For [user], [product] is the only [category] that [claim] because [reasoning]" format}

## Core Features

| # | Feature | Why Required for USP | Simplest Implementation |
|---|---------|---------------------|------------------------|
| 1 | ... | ... | ... |
| 2 | ... | ... | ... |
| 3 | ... | ... | ... |

## Known Risks

| Category | Risk | Grade | Notes |
|----------|------|-------|-------|
| Market | ... | [KNOWN] or [ASSUMED] | ... |
| Technical | ... | [KNOWN] or [ASSUMED] | ... |
| Legal | ... | [KNOWN] or [ASSUMED] | ... |
| Distribution | ... | [KNOWN] or [ASSUMED] | ... |

## Target User

{2–3 sentence persona sketch}

## Success Criteria

1. {Measurable criterion with target metric and timeframe}
2. {Measurable criterion with target metric and timeframe}

## Phase 1 Research Targets

{List of [ASSUMED] items from the risks and claims above that Phase 1 agents should validate}

Validation Checklist

• USP hypothesis is a single falsifiable sentence
• Core features are ≤ 5 and each justifies its existence via the USP
• Every risk is tagged [KNOWN] or [ASSUMED]
• Success criteria are measurable (include numbers and timeframes)
• Grade is set to hypothesis and confidence to 0.3
• Phase 1 research targets are listed for every [ASSUMED] item