lope-deliberate

name

lope-deliberate

description

Run an Agent-Order-style council deliberation on a decision artifact. Six built-in templates: ADR, PRD, RFC, build-vs-buy, migration-plan, incident-review. The 7-stage protocol — independent positions, anonymized critique, revision, synthesis, rubric review, minority report — turns a scenario file into a structured artifact that has survived peer scrutiny. Use when the user asks 'should we adopt X?', 'review this ADR / PRD / RFC', 'build vs buy this capability?', 'plan the migration', 'run the incident review'. Pure adversarial reasoning — no source-file modification, no git mutations.

Lope Deliberate

lope deliberate <template> <scenario> runs the council on a decision artifact. The output is a structured directory with the synthesis, a minority report, a decision log, and a redacted JSONL trace of every turn.

Invocation

Two paths — you must handle both:

1. Explicit slash command / direct CLI. User types lope deliberate adr scenario.md or lope deliberate prd specs/x.md --depth quick.

2. Natural language. User says one of:

   • "Help me decide whether to adopt JWT auth" → lope deliberate adr <scenario>
   • "Review this ADR / PRD / RFC" → lope deliberate <template> <file>
   • "Build vs buy our secrets vault" → lope deliberate build-vs-buy <file>
   • "Plan the Postgres migration" → lope deliberate migration-plan <file>
   • "Run the incident review for last Friday" → lope deliberate incident-review <file>

   When the user supplies the prose but no file, ask them to drop the scenario into a markdown file (or pipe via lope deliberate <template> -).

Templates

Name	Title	When to use
adr	Architecture Decision Record	"Should we adopt X for our architecture?"
prd	Product Requirements Document	"Define what we're building before we build it."
rfc	Request for Comments	"Detailed design open for council critique."
build-vs-buy	Build vs Buy Analysis	"Make or buy this capability?" — 3-year TCO horizon.
migration-plan	Migration Plan	"Cut over from A to B with rollback at every phase."
incident-review	Incident Review	"Postmortem with timeline, root cause, owned action items."

Command shape

# Default: standard depth, anonymous critique, output to lope-runs/<ts>-<template>/
lope deliberate adr scenario.md

# Quick mode skips critique + revision rounds
lope deliberate adr scenario.md --depth quick

# Deep mode preserves every dissent in the minority report
lope deliberate prd scenario.md --depth deep

# Force a specific output directory
lope deliberate rfc scenario.md --out custom-runs/jwt-rfc/

# Postmortem mode: validator names visible (de-anonymize)
lope deliberate incident-review incident.md --no-anonymize

# Pipe a scenario from stdin
cat scenario.md | lope deliberate adr -

# Brain-aware deliberation (Makakoo OS only)
lope deliberate adr scenario.md --brain-context "prior auth decisions" --brain-log

What gets written

lope-runs/<timestamp>-<template>/
  scenario.md           # input copy (redacted)
  trace.jsonl           # per-turn record (redacted; validator names hidden in anonymous mode)
  turns/
    01-positions/*.md
    02-critiques/*.md
    03-revisions/*.md
  final/
    report.md           # synthesis with template sections
    minority-report.md  # preserved dissent (or "unanimous" note)
    decision-log.md     # template + council size + verdicts + timestamps

Hard rules

• No source-file modification. The verb is read-only against the user's tree. Tests pin this.
• No git mutation. No commit, no branch, no stash.
• Anonymous by default. Validator names never reach the critique prompt unless --no-anonymize is passed. The same validator gets the same Response X label across every stage (positions, critiques, revisions, synthesis source list, rubric, minority report).
• Always emit a minority report. Even when the rubric is unanimous PASS the file is written with a "no dissent" note so downstream scripts have a stable shape.
• Trace JSONL is redaction-clean. Every turn text + metadata passes through redact_text.

Depth presets

• --depth quick — skip stages 3 (critique) and 4 (revision). Use when the user wants a fast first-cut artifact rather than a peer-reviewed one.
• --depth standard (default) — full 7-stage protocol; minority report keeps high-severity NEEDS_FIX objections.
• --depth deep — full protocol; minority report preserves every dissent, not just high-severity.

Pairing with other v0.7 verbs

• --brain-context "..." and --brain-log plug into Makakoo OS Brain.

For full details: lope docs (docs/deliberation.md).

Runtime safety (v0.14.0)

Use --run-timeout for the whole command; --timeout remains a per-provider-call ceiling. Inspect the emitted request plan before large work and prefer compact evidence. Automatic shaping is bounded; use chunking only when the forecast fits --max-calls and --max-chunks. For abandoned work, run lope jobs list and preview lope jobs reap --dry-run. Never use pkill -f, killall, or process-name matching.