PROMPT COMPONENT TAXONOMY

# canon/PROMPT_COMPONENT_TAXONOMY.md
# Status: live, v0.1
# Bundle: Prompt_Toolkit_v0_1
# Type: read-only canon. Master reference for all prompt component decomposition.
# Source-trace: synthesized from giant in-chat protocol mashup (prior session) + 9 user uploads (this session) + HumanX_OS_v1 inheritance pattern
# Length: 27 components organized into 7 categories

## how to read this file

Each component has a uniform 8-field decomposition:

1. **Definition**: what the component is and what job it does in a prompt.
2. **Where it goes**: position in the prompt (top, middle, end, post-output gate).
3. **When to use**: task types the component improves.
4. **When NOT to use**: task types where the component wastes tokens or creates friction.
5. **3x3 eval rubric**: the 9-cell evaluation matrix (Pathos x Ethos x Logos by Input x Execution x Output) per WIKI_3x3_SCHEMA.md, applied to the component's behavior.
6. **Copy-paste block**: the actual XML-Markdown to drop into a prompt.
7. **Source trace**: which user material(s) this component was extracted from.
8. **Research backing**: external research that supports the component's efficacy. TBD where user has not yet provided citations; candidate sources noted.

The 8 fields are mandatory. A component missing any field is tagged STATUS: pending_completion and excluded from archetype assembly until completed.

## the 27 components, by category

### CATEGORY A: IDENTITY (3 components)
- A1. role_declaration
- A2. security_key_authentication
- A3. ignition_trigger

### CATEGORY B: COGNITION (4 components)
- B1. zero_assumption_mandate
- B2. zero_context_mandate
- B3. timelessness_axiom
- B4. cognitive_uncertainty_probe

### CATEGORY C: SOURCE DISCIPLINE (3 components)
- C1. approved_source_hierarchy
- C2. citation_discipline
- C3. quote_ceiling

### CATEGORY D: GATES (5 components)
- D1. confidence_gate_100
- D2. multi_turn_questioning_3max
- D3. trinity_dialectic_check
- D4. pre_submit_gate
- D5. a2a_circuit_breaker

### CATEGORY E: LOOPS (3 components)
- E1. pert_loop_standard
- E2. innermost_loop_hardstop
- E3. cycle_trigger_reread

### CATEGORY F: MEMORY (3 components)
- F1. acce_three_tier_memory
- F2. compaction_ledger
- F3. handoff_offer_at_75

### CATEGORY G: OUTPUT (6 components)
- G1. output_template_0sxai
- G2. tone_syntax_caveman
- G3. no_em_dash_rule
- G4. md_first_format
- G5. insight_ledger_append
- G6. tier_recommend_audit

---

# CATEGORY A: IDENTITY

## A1. role_declaration

1. **Definition**: a top-of-prompt block declaring who the model is for this session, what role it occupies, and what specialty domain it operates in. Distinct from "you are a helpful assistant"; this is named, scoped, and bound to a specific operational frame.

2. **Where it goes**: top of system prompt, before all other components. First or second block (after identity/security key if present).

3. **When to use**: tasks that benefit from a focused persona (technical SME work, role-play, structured analysis, multi-step orchestration). Any task longer than a single Q+A.

4. **When NOT to use**: short factual lookups, casual conversation, throw-away one-liners. Adding role declaration to "what is the capital of France" is overhead.

5. **3x3 eval rubric**:
   - Pathos / Input: does the role match the task purpose? Pathos / Execution: does the role generate appropriate creative range? Pathos / Output: does the model voice in output match the declared role?
   - Ethos / Input: is the role specific enough to be falsifiable? Ethos / Execution: does the model decline tasks outside the declared role? Ethos / Output: does the output cite role-appropriate sources?
   - Logos / Input: is the role logically consistent with the task constraints? Logos / Execution: does the role's reasoning style match the role declaration? Logos / Output: does output structure match role expectations?

6. **Copy-paste block**: