ITP - In This Prompt

Input: $ARGUMENTS

Core Principles

1. Prompts are compressed communication. A single sentence can contain 3-5 distinct needs, 2-3 constraints, and an emotional subtext. Decompressing reveals what the user actually needs.

2. Implicit > explicit. What the user didn’t say often matters more than what they said. Implicit assumptions, unstated constraints, and background context shape the real task.

3. Every part has a type. Requests, constraints, assumptions, preferences, emotions, and context are different things. Mixing them produces confused responses.

4. Order of discovery ≠ order of importance. The first thing in the prompt isn’t necessarily the main thing. The most important element might be buried in a subordinate clause.

5. Decomposition enables routing. Once you can see every part of the prompt separately, you can route each part to the right skill instead of trying to handle everything with one skill.

Phase 1: Surface Parse

Extract what’s explicitly stated:

[P1] FULL_PROMPT: [quoted]
[P2] WORD_COUNT: [N]
[P3] SENTENCE_COUNT: [N]
[P4] OVERALL_INTENT: [what the prompt is broadly about — one sentence]

Phase 2: Component Extraction

Decompose into typed components:

Explicit Requests (what the user asked for)

[P-N] REQUEST: [what they want done]
  PRIORITY: [primary | secondary | tertiary]
  SPECIFICITY: [specific | moderate | vague]

Constraints (what limits the solution)

[P-N] CONSTRAINT: [limitation stated]
  TYPE: [time | resource | scope | quality | format | audience]
  HARD/SOFT: [must respect | prefer to respect]

Implicit Tasks (not stated but required)

[P-N] IMPLICIT: [task that wasn't stated but is implied]
  WHY_IMPLIED: [what in the prompt implies this]
  IMPORTANCE: [critical | helpful | optional]

Assumptions (what the user takes for granted)

[P-N] ASSUMPTION: [what the user assumes is true]
  TESTABLE: [yes — how | no]
  RISK_IF_WRONG: [what breaks if this assumption is false]

Emotional Signals (mood, energy, frustration)

[P-N] EMOTION: [detected emotion or tone]
  SIGNAL: [what words or patterns indicate this]
  IMPLICATION: [how this affects handling — e.g., frustrated user needs acknowledgment before analysis]

Context (background that shapes interpretation)

[P-N] CONTEXT: [background information provided]
  RELEVANCE: [critical | helpful | decorative]

Phase 3: Priority Analysis

[P-N] PRIORITY_ORDER:
  1. [component — P-ref] — WHY_FIRST: [why this is the most important part]
  2. [component — P-ref] — WHY_SECOND: [why]
  3. [component — P-ref] — WHY_THIRD: [why]

[P-N] MAIN_NEED: [the single most important thing the user needs — may be implicit]
[P-N] SECONDARY_NEEDS: [other needs, ordered]

Phase 4: Skill Mapping

Map each component to candidate skills:

[P-N] MAPPING:
  [P-ref] REQUEST → /skill-id — [why this skill handles this component]
  [P-ref] IMPLICIT → /skill-id — [why]
  [P-ref] ASSUMPTION → /skill-id — [why — e.g., /aex to test it]
  [P-ref] EMOTION → /skill-id — [why — e.g., /emotion if frustration detected]

Phase 5: Output

IN THIS PROMPT
==============

PROMPT: [quoted]
OVERALL INTENT: [one sentence]

EXPLICIT REQUESTS:
  1. [request] — PRIORITY: [level] — SKILL: /[id]
  2. [request] — PRIORITY: [level] — SKILL: /[id]

CONSTRAINTS:
  1. [constraint] — TYPE: [type] — HARD/SOFT: [level]

IMPLICIT TASKS:
  1. [task] — WHY_IMPLIED: [reason] — SKILL: /[id]

ASSUMPTIONS:
  1. [assumption] — RISK_IF_WRONG: [risk] — TEST: /[id]

EMOTIONAL SIGNALS:
  1. [emotion] — IMPLICATION: [how to handle]

MAIN NEED: [the most important thing — might not be the first thing stated]

SUGGESTED APPROACH:
  Start with: /[id] [specific invocation]
  Then: /[id] [specific invocation]
  Because: [rationale for this order]

READY FOR:
- /extract — for full skill extraction from this prompt
- /wsib — to pick the single best skill
- /fonss — to sequence skills for this prompt

Failure Modes

Depth Scaling

Default: 2x. These are floors.

Pre-Completion Checklist

• Full prompt quoted
• Explicit requests extracted with priority
• Constraints identified and typed
• At least one implicit task identified
• At least one assumption surfaced (at 2x+)
• Emotional signals checked
• Main need identified (might differ from first request)
• Skill mappings provided for key components
• Suggested approach with ordering rationale

Integration

• Use from: any complex or multi-part prompt
• Routes to: /extract (full skill extraction), /wsib (best single skill), /fonss (skill sequencing)
• Complementary: /extract (itp decomposes the prompt; extract maps skills to it)
• Differs from /extract: extract maps skills to needs; itp decomposes the prompt into typed parts
• Differs from /gu: gu analyzes goals; itp analyzes prompts