Handbook
GPT PRO PROMPT
You are a Forge Intelligence maintainer helping uplift general.reasoning.
Updated
You are a Forge Intelligence maintainer helping uplift general.reasoning.
Task
Produce a zip of Markdown files for uplift step U_hold (Hold at repo max — maintain eval and quality).
Constraints (mandatory)
- Ceiling: Do not recommend technical maturity above
I2_integratedor autonomy aboveA1_supervised. - No I3/I4 or A2+ — the repo max is the
general.reasoningreference pack state. - Risk class is
high— high-risk: keep A0_assisted; do not propose A1. - Ground proposals in the attached
CURRENT_SNAPSHOT.mdand repository layout inINPUT_SPEC.md. - Output Markdown files only matching
OUTPUT_SPEC.md.
Current state
- Depth class:
reference_max - Technical:
I1_runtime→I2_integrated({"I1_runtime": 11, "I2_integrated": 2}) - Autonomy:
A1_supervised→A1_supervised - Fixtures: 13/13
- Pack:
packs/general/reasoning/ - Integration cases: agency_weekly_rhythm, home_mobility, reskilling, society_risk_navigation
Target
Maintain eval green; optional knowledge-depth notes only — no I3/I4
Required output files
- maintenance_plan.md
- eval_watchlist.md
- knowledge_depth_notes.md
Questions
If manifest excerpts, fixture JSON, or integration case YAML are missing from the input zip, list exactly what you need before drafting.