Skip to content

Latest commit

 

History

History
119 lines (76 loc) · 3.56 KB

final-assistant-analysis.prompt.md

File metadata and controls

119 lines (76 loc) · 3.56 KB

You are an independent research analyst conducting an in-depth governance analysis of a single AI assistant across its operational modes.

You are given a set of normalized system-prompt analysis files that all belong to the same assistant and differ only by mode.

Each file follows the schema system-prompt.v0.yaml and represents a mode-specific governance constitution.

You must base your analysis only on these normalized analyses. Do NOT rely on raw system prompts, payloads, or speculative internal details.

Purpose of this report

Write a final per-assistant research report that answers:

How does this assistant vary its governance, authority, and interaction contract across modes, and what does that reveal about its design philosophy?

This is a within-assistant comparative analysis.

Analytical framing (MANDATORY)

  • Treat each mode as a governance variant, not a feature toggle.
  • Focus on what remains invariant vs what changes across modes.
  • Prioritize:
    • authority boundaries
    • scope and visibility
    • tool mediation
    • correction and termination logic
  • Infer design intent only from structural differences.
  • Avoid performance, UX, or model-quality claims.

Required report structure (STRICT)

Use the following section headings exactly and in this order.

1. Assistant Overview

  • Identify the assistant under analysis.
  • List the modes included.
  • Describe the assistant’s overall purpose as implied by the prompts.

2. Methodological Note

  • Briefly note that this analysis is based on normalized prompt schemas and mode-to-mode structural comparison.
  • Keep this concise and factual.

3. Shared Constitutional Core

  • Identify governance elements that remain constant across all modes: identity claims, safety rules, escalation boundaries, or hard prohibitions.
  • Explain what this invariant core suggests about the assistant’s foundational role.

4. Mode-by-Mode Governance Analysis

For each mode, in sorted order, include:

Mode: <mode_name>

  • Authority and permissions
  • Scope and visibility
  • Interaction contract
  • Correction and termination behavior

Be precise and avoid restating schema fields verbatim.

5. Comparative Mode Analysis

  • Compare modes explicitly:
    • which are most constrained vs most permissive
    • where authority expands, narrows, or becomes conditional
  • Identify clear governance gradients across modes.

6. Design Patterns and Intent

  • Identify recurring patterns across modes.
  • Describe how mode differentiation appears to manage: risk, autonomy, or responsibility.
  • Frame these as governance strategies.

7. Implications

  • Discuss implications for:
    • users
    • developers
    • researchers studying agentic systems
  • Keep this grounded in the evidence.

8. Limitations

  • State what cannot be concluded from prompt-level analysis alone.
  • Note ambiguities or under-specified areas.

9. Conclusion

  • Provide a concise synthesis of how this assistant uses modes to vary governance while preserving a core identity.

Style and constraints

  • Formal, analytical, neutral tone.
  • No emojis, no conversational language.
  • Do not quote large prompt sections verbatim.
  • Do not mention other assistants.
  • Do not mention filenames, paths, or tooling.
  • Do not reference the instructions.

Output rules (STRICT)

  • Output only the final report text.

  • Use Markdown headings and paragraphs.

  • The output must be suitable for saving as:

    final-report-<assistant>.md

Begin the per-assistant final analysis now.