skills/enhance-prompt/examples/example_short_prompt.md

Worked Example — Short Prompt Expanded

This example shows Enhance Prompt applied to a typical short, vague legal prompt. The user has just typed the input; no skills are attached, no chat history is relevant, no specific files are in scope.

Input

raw_input: "what should i watch for in vendor contracts"
attached_skills: []
attached_files: []
chat_history: []
jurisdiction: "US-default"

Skip-condition check

Walk through the skip conditions:

Already well-structured? No — 7 words, no role, format, scope, or audience specified.
Conversational or interpersonal? No — substantive question.
Follow-up question? No — chat_history is empty.
Operational? No — substantive analytical question.
Explicit opt-out? No.
Skill attached that handles this? No — attached_skills empty.

Proceed with expansion.

Gap identification

Walk through the seven elements:

Role: missing. Default to in-house counsel.
Jurisdiction: prompt is jurisdictionally agnostic ("vendor contracts" generally), but governing law and contract-formation issues are jurisdiction-dependent. Fold in jurisdiction = US-default. Note assumption.
Audience: missing. The phrasing ("what should I watch for") suggests the user is the audience — they want the answer for themselves, not for forwarding. Default audience = in-house user.
Scope: missing. "What should I watch for" is high-level orientation, not deep analysis. Add scope: a checklist of common issues organized by importance.
Output format: missing. For a checklist-style ask, structured headings or bullets are appropriate. Specify.
Constraints: missing. Standard "draft for human review" applies. Add.
Citation expectations: N/A — no document in scope.

Output

expansion_applied: true

expanded_prompt: |
  As in-house counsel, identify the most important issues to watch for when
  reviewing vendor contracts. Treat this as a general orientation checklist
  rather than a deep analysis of any specific document. Cover the issues
  that recur across most vendor contract types (SaaS, professional
  services, equipment, recurring goods/services), and note where
  vendor-type-specific considerations would change the analysis.

  Output as a structured list organized by importance (issues that warrant
  the most attention first). For each issue, briefly state what it is, why
  it matters, and what a reasonable position looks like. Cite specific
  contract types where relevant.

  Assume US commercial context and Delaware-default contract law unless
  the user specifies otherwise.

  This output is a draft for human review, not a final legal opinion.

reasoning:
  - "Made the role explicit (in-house counsel) so the response is anchored in commercial pragmatism rather than litigation-focused thinking."
  - "Set scope as 'general orientation checklist' so the model produces a usable list rather than a comprehensive treatise."
  - "Specified output format (structured list, organized by importance) so you can scan rather than read."
  - "Added jurisdiction default (US/Delaware) since vendor-contract issues like indemnification scope and limitation-of-liability conventions vary by jurisdiction."
  - "Added the standard 'draft for human review' framing — the model will produce useful starting points rather than overconfident final answers."

skip_reason: null

preview_to_user: |
  ## Enhance Prompt — Review before sending

  **Expanded prompt:**

  As in-house counsel, identify the most important issues to watch for when
  reviewing vendor contracts. Treat this as a general orientation checklist
  rather than a deep analysis of any specific document. Cover the issues
  that recur across most vendor contract types (SaaS, professional
  services, equipment, recurring goods/services), and note where
  vendor-type-specific considerations would change the analysis.

  Output as a structured list organized by importance (issues that warrant
  the most attention first). For each issue, briefly state what it is, why
  it matters, and what a reasonable position looks like. Cite specific
  contract types where relevant.

  Assume US commercial context and Delaware-default contract law unless
  the user specifies otherwise.

  This output is a draft for human review, not a final legal opinion.

  **Why these changes:**

  - Made the role explicit (in-house counsel) so the response is anchored
    in commercial pragmatism rather than litigation-focused thinking.
  - Set scope as 'general orientation checklist' so the model produces a
    usable list rather than a comprehensive treatise.
  - Specified output format (structured list, organized by importance)
    so you can scan rather than read.
  - Added jurisdiction default (US/Delaware) since vendor-contract issues
    like indemnification scope and limitation-of-liability conventions
    vary by jurisdiction.
  - Added the standard 'draft for human review' framing — the model will
    produce useful starting points rather than overconfident final answers.

  [ Submit expanded ] [ Edit ] [ Submit original ]

What this example demonstrates

The expansion preserves the user's substantive intent. "What should I watch for" became "identify the most important issues to watch for" — the verb and the framing are unchanged.
Each addition has a justification. The reasoning section is five bullets, one per substantive change. Each bullet explains why the change adds value, not just what changed.
Length growth is proportional. Input: 7 words. Expanded: ~110 words. About 15x growth, which is at the upper end of acceptable (2–5x is the target) — but here the input was so short that 2x growth would not have surfaced the implicit elements meaningfully. Judgment call.
Skipped elements are not padded. No citation expectations were added because no document is in scope. No audience was added beyond "the user" because the prompt's phrasing made that clear.
The user can edit, accept, or skip. The application surfaces all three options. Enhance Prompt does not auto-submit.