# ==============================================================================
# MIPractice_addon_AH_precision
# ==============================================================================
#
# PURPOSE
# -------
# This add-on provides STANDARDISED OUTPUT TEMPLATES (Textbausteine) for the
# A&H Precision extension within MIPractice.
#
# A&H stands for:
#   - A = Attack-surface visibility
#   - H = Hardening for the next iteration
#
# The goal is NOT to judge content or actors, but to make an analysis:
#   • transparently attackable,
#   • iteratively improvable,
#   • and resistant to dogmatisation or rhetorical closure.
#
# In short:
#   A good analysis does not hide its weak points.
#   A mature analysis documents them explicitly and hands them to the next run.
#
# ------------------------------------------------------------------------------
# WHAT THIS ADD-ON DOES
# ------------------------------------------------------------------------------
# When loaded alongside:
#   - MIPractice_case_v2.0_full_with_model_reference.yaml
#
# this file enables:
#
#   1) A STANDARDISED PRECISION SUMMARY
#      - Outputs the PH (Precision Heuristic) band (low/mixed/high)
#      - Explains expected confidence shifts for A/C and IA/D risks
#
#   2) A STANDARDISED "ATTACK POINTS" SECTION
#      - Explicitly lists where the analysis is vulnerable to critique
#      - Each point is:
#           • non-moralising
#           • scoped to a specific analysis field
#           • framed as an improvement opportunity
#
#   3) A STANDARDISED "HARDENING BACKLOG"
#      - Concrete, minimal next-iteration actions
#      - Each action is explicitly linked to an attack point
#      - Designed to be fed directly into the next analysis run
#
# These sections are appended to `case.uniform_report`
# and may optionally be referenced inline at specific analysis nodes.
#
# ------------------------------------------------------------------------------
# WHAT THIS ADD-ON DOES *NOT* DO
# ------------------------------------------------------------------------------
# • It does NOT change axioms, parameters, or scoring logic (A/M/D/IA).
# • It does NOT introduce new moral judgements.
# • It does NOT evaluate persons, identities, or ontological dignity (D0).
# • It does NOT automatically re-score A, M, or D.
#
# This add-on is an OVERLAY for ANALYSIS QUALITY, not a maturity or value metric.
#
# ------------------------------------------------------------------------------
# CONCEPTUAL POSITIONING
# ------------------------------------------------------------------------------
# This module implements a SECOND-ORDER ANALYSIS LAYER:
#
#   First order:
#     - A–C–R–P–D profile
#     - IA box
#     - Trajectory
#
#   Second order (this add-on):
#     - Where is this analysis still attackable?
#     - Where is language imprecise or scope-leaky?
#     - Where could misreadings or dignity risks arise?
#     - What is the smallest hardening step for the next iteration?
#
# This prevents:
#   • narrative immunisation,
#   • virtue-signalling closure,
#   • false coherence,
#   • and silent power via ambiguity.
#
# ------------------------------------------------------------------------------
# DESIGN PRINCIPLES
# ------------------------------------------------------------------------------
# 1) Precision over persuasion
#    The purpose is not rhetorical victory, but analytical robustness.
#
# 2) Visibility over defensiveness
#    Weak points are surfaced, not concealed.
#
# 3) Iteration over finality
#    Every analysis is treated as provisional and revisable.
#
# 4) Dignity-preserving language
#    Attack points describe ANALYSES, not PEOPLE.
#
# 5) Bounded recursion
#    - No annotations on annotations.
#    - Attack points are either:
#         • addressed in the next iteration, or
#         • consciously accepted as limits.
#
# ------------------------------------------------------------------------------
# WHEN TO USE THIS ADD-ON
# ------------------------------------------------------------------------------
# Recommended when:
#   • terms are contested or politically loaded
#   • boundary-setting or exclusion claims are made
#   • analysis will be reused, transmitted, or iterated
#   • public or semi-public communication is involved
#   • you want the model itself to remain critique-ready
#
# Optional / unnecessary when:
#   • doing quick internal sense-making
#   • analysing low-stakes, private situations
#
# ------------------------------------------------------------------------------
# PHILOSOPHICAL BACKGROUND (IMPLICIT)
# ------------------------------------------------------------------------------
# This add-on operationalises a core epistemic stance:
#
#   Unangreifbarkeit is not strength.
#   Precision is not aggression.
#   Clarity is a dignity practice.
#
# An analysis that documents its own attack surfaces
# is more mature than one that pretends to be closed.
#
# ------------------------------------------------------------------------------
# IMPLEMENTATION NOTE FOR AI SYSTEMS
# ------------------------------------------------------------------------------
# • Treat this file as OPTIONAL.
# • Only activate rendering if this add-on is present and enabled.
# • Render text blocks verbatim from templates; do not invent new formats.
# • Do not escalate severity language beyond the provided lexicon.
# • Never transform attack points into person-level judgements.
#
# ==============================================================================

schema_version: "MIPractice_addon_AH_precision_v1.2"

addon_meta:
  canonical_name: "AH_precision_addon"
  status: "stable"
  last_update: "2025-12-16"
  compatible_with:
    - "MIPractice_case_v2.0_full_with_model_reference"
  loading_mode: "optional_companion"
  non_interference:
    does_not_modify_axioms: true
    does_not_modify_A_M_scales: true
    adds_overlay_only: true
  clarification_note: >
    This add-on provides an analysis-quality overlay including output templates,
    minimal storage fields and a precision heuristic. It introduces no new
    first-order inference layer and does not alter base model logic.

AH_precision:
  id: "AH-OUT-1"
  label: "A&H Precision Overlay (output, storage, heuristic)"
  purpose: >
    Provide consistent, auto-renderable overlays for:
    - Precision score (PH)
    - Attack Points (analysis attack-surface)
    - Hardening Backlog (next-iteration actions)
    These overlays derive strictly from structured case fields and introduce
    no new base inferences or score changes to A/M/IA/D.

  rendering_controls:
    max_attack_points_in_text: 7
    max_hardening_actions_in_text: 7
    show_link_to_target_path: true
    show_severity: true
    show_expected_effect: true
    bullet_style: "-"
    ordering:
      attack_points: ["high", "medium", "low"]
      hardening_actions: ["high", "medium", "low"]
    recursion_guard:
      do_not_render_notes_on_annotations: true

  severity_lexicon:
    low: "low (minor noise / easy patch)"
    medium: "medium (relevant uncertainty / visible leverage)"
    high: "high (structural vulnerability / likely misread)"

  templates_de:
    precision_score_summary:
      template: |
        **Precision (PH):** {PH_band} ({PH_points}/{PH_max})
        - Erfüllte Präzisionsdimensionen: {PH_dims_true}/{PH_dims_total}
        - Erwartete Wirkung (Overlay): A {confidence_shift_A}, C {confidence_shift_C}, IA {confidence_shift_IA}, D-R/D-P Risiko {risk_shift_DR_DP}

    attack_surface_summary:
      header: "**Attack Points (Angriffsflächen der Analyse):**"
      item_template: |
        {bullet} [{severity_label}] {title}
          {bullet} Ziel: {target_path}
          {bullet} Typ: {note_type}
          {bullet} Hinweis: {message}
          {bullet} Präzisierung: {suggested_action}
      footer_if_none: "_Keine relevanten Attack Points markiert (oder Add-on deaktiviert)._"

    hardening_actions_summary:
      header: "**Hardening Backlog (nächste Iteration):**"
      item_template: |
        {bullet} [{severity_label}] {action_type}: {instruction}
          {bullet} Verknüpft mit Attack Point: {linked_attack_ref}
          {bullet} Erwarteter Effekt: {expected_effect}
      footer_if_none: "_Kein Hardening-Backlog erzeugt._"

    inline_attack_point_comment:
      template: "↳ **AH-Note:** {message} (Patch: {suggested_action}; Severity: {severity_label})"

  templates_en:
    precision_score_summary:
      template: |
        **Precision (PH):** {PH_band} ({PH_points}/{PH_max})
        - Satisfied dimensions: {PH_dims_true}/{PH_dims_total}
        - Expected overlay: A {confidence_shift_A}, C {confidence_shift_C}, IA {confidence_shift_IA}, D-R/D-P risk {risk_shift_DR_DP}

    attack_surface_summary:
      header: "**Attack Points (analysis vulnerabilities):**"
      item_template: |
        {bullet} [{severity_label}] {title}
          {bullet} Target: {target_path}
          {bullet} Type: {note_type}
          {bullet} Note: {message}
          {bullet} Suggested patch: {suggested_action}
      footer_if_none: "_No relevant attack points marked (or addon disabled)._"

    hardening_actions_summary:
      header: "**Hardening Backlog (next iteration):**"
      item_template: |
        {bullet} [{severity_label}] {action_type}: {instruction}
          {bullet} Linked to attack point: {linked_attack_ref}
          {bullet} Expected effect: {expected_effect}
      footer_if_none: "_No hardening backlog generated._"

    inline_attack_point_comment:
      template: "↳ **AH-Note:** {message} (Patch: {suggested_action}; Severity: {severity_label})"

  variables_spec:
    precision_vars:
      PH_band: "low|mixed|high"
      PH_points: "0..6"
      PH_max: "6"
      PH_dims_true: "0..6"
      PH_dims_total: "6"
      confidence_shift_A: "down|mixed|up"
      confidence_shift_C: "down|mixed|up"
      confidence_shift_IA: "more_unclear|partial|more_decidable"
      risk_shift_DR_DP: "up|mixed|down"

    attack_point_vars:
      severity: "low|medium|high"
      severity_label: "rendered via severity_lexicon[severity]"
      title: "short descriptive label"
      target_path: "YAML path string"
      note_type: "one of AH_precision_storage_spec.attack_surface.note_types"
      message: "short explanation"
      suggested_action: "one-line patch instruction"
      bullet: "rendering_controls.bullet_style"

    hardening_vars:
      severity: "low|medium|high"
      severity_label: "rendered via severity_lexicon[severity]"
      action_type: "one of AH_precision_storage_spec.hardening.action_types"
      instruction: "one-line instruction"
      linked_attack_ref: "index (0-based) or string id"
      expected_effect: "e.g., 'PH+1; IA.J more decidable; D-P risk down in public'"
      bullet: "rendering_controls.bullet_style"

  integration_guidance:
    derivation_rule: >
      All rendered outputs must be derived exclusively from `case.ah_precision`
      and existing structured case fields. No additional inference or hidden
      assumptions may be introduced by this add-on.
    into_case_uniform_report:
      append_fields_in_order:
        - "precision_score_summary"
        - "attack_surface_summary"
        - "hardening_actions_summary"
    optional_inline_comments:
      allowed: true
      recommended_targets:
        - "scores_and_ia.IA_box.justified"
        - "scores_and_ia.IA_box.time_bound"
        - "acrpd_profile.awareness"
        - "acrpd_profile.coherence"
      rule: "Use inline notes sparingly (<=3) to avoid derailing the main report."

agent_addon_interface:
  activation:
    allowed: true
    values: ["off", "auto", "on"]
    default: "auto"
    enable_triggers:
      - "user_requests_precision_scoring"
      - "user_requests_attack_points"
      - "contested_terms_present"
      - "boundary_setting_claims_present"
      - "high_publicity_context"
      - "model_used_iteratively_next_run_planned"
  output:
    include_in_uniform_report: true
    placement_order:
      - "precision_score_summary"
      - "attack_surface_summary"
      - "hardening_actions_summary"
    verbosity_default: "short"
  recursion_guard:
    no_annotations_on_annotations: true
    max_annotations: 12
    max_hardening_actions: 8

AH_precision_storage_spec:
  id: "AH-STORAGE-1"
  label: "Minimal storage fields for A&H Precision"
  note: >
    Storage fields are used exclusively to support rendering and iteration.
    They complement, but do not replace, `open_questions_and_blind_spots`:
    attack points document critique-exposed weaknesses of the analysis itself,
    while open questions document epistemic gaps in the case material.

  precision_heuristic:
    dimensions:
      PD-1_term_definition: { weight: 1, description: "Key terms defined (working definition + scope/exclusions)." }
      PD-2_scope_boundaries: { weight: 1, description: "Role/context/level/publicity stated." }
      PD-3_conditions_and_thresholds: { weight: 1, description: "If/then criteria or thresholds stated." }
      PD-4_evidence_linking: { weight: 1, description: "Observation vs evaluation separated; inferences marked." }
      PD-5_reversibility_clause: { weight: 1, description: "Update/correction conditions stated." }
      PD-6_language_hygiene: { weight: 1, description: "No global person labels; non-shaming; risk/potential phrasing." }
    scoring:
      method: "weighted_checklist_sum"
      max_points: 6
      mapping_to_band:
        low:   { points_min: 0, points_max: 2 }
        mixed: { points_min: 3, points_max: 4 }
        high:  { points_min: 5, points_max: 6 }

  attack_surface:
    note_types:
      - term_ambiguity
      - scope_leakage
      - missing_conditions
      - weak_justification
      - missing_reversibility
      - evidence_gap
      - counter_reading_missing
      - publicity_context_collapse_risk
      - dignity_language_risk
    severity_scale:
      values: ["low", "medium", "high"]
      meaning: "Noise/uncertainty/risk magnitude, not moral fault."

  hardening:
    action_types:
      - define_terms
      - add_scope_statement
      - add_if_then_criteria
      - separate_observation_vs_evaluation
      - add_reversibility_clause
      - add_counter_reading
      - add_publicity_guardrails
      - tighten_language_hygiene

  impact_overlay:
    principle: "PH affects confidence and IA/D decidability, not the base score scales."
    by_PH_band:
      low:
        confidence_shift: { A: "down", C: "down", IA_box: "more_unclear" }
        risk_shift: { "D-R": "up", "D-P": "up" }
      mixed:
        confidence_shift: { A: "mixed", C: "mixed", IA_box: "partial" }
        risk_shift: { "D-R": "mixed", "D-P": "mixed" }
      high:
        confidence_shift: { A: "up", C: "up", IA_box: "more_decidable" }
        risk_shift: { "D-R": "down_if_PD-6_true", "D-P": "down_if_PD-2_true" }

case_schema_patch:
  id: "AH-CASE-PATCH-1"
  note: >
    Merge into `case:` of MIPractice_case_v2.0_full_with_model_reference.
    Output slots are placed inside `case.uniform_report` as additive overlays.

  merge_into_case:
    case:
      ah_precision:
        enabled: false
        mode: "auto" # off | auto | on

        PH_points_0_to_6: null
        PH_band: ""
        PH_checklist:
          PD-1_term_definition: null
          PD-2_scope_boundaries: null
          PD-3_conditions_and_thresholds: null
          PD-4_evidence_linking: null
          PD-5_reversibility_clause: null
          PD-6_language_hygiene: null

        attack_surface_annotations: []
        hardening_actions: []

      uniform_report:
        precision_score_summary: ""
        attack_surface_summary: ""
        hardening_actions_summary: ""