hashgraph-online/awesome-codex-plugins / plugins/beastlabai/multi-llm-plugin/skills/multi-llm
multi-llm
FlaggedOrchestrate plan reviews, task generation, implementation and code reviews across several LLM CLIs in parallel, or ask every model one free-text question about a plan. Use when the user says: review my plan, multi-llm, run a code review, generate tasks from a plan, implement this plan, ask the models about my plan.
display_decisions.py
View on GitHub#!/usr/bin/env python3
"""
Display a formatted summary of suggestion/fix decisions from orchestrator output.
Reads orchestrator_output.json and shows which items will be applied, skipped,
need human review, etc., with their titles.
Usage:
uv run --project "{SKILL_DIR}" -- python "{SKILL_DIR}/multi_llm.py" util decisions \
--output-file plans/my-feature/apply-suggestions/orchestrator_output.json
# Or auto-detect from plan path + phase:
uv run --project "{SKILL_DIR}" -- python "{SKILL_DIR}/multi_llm.py" util decisions \
--plan-file plans/my-feature.md --phase apply-suggestions
"""
import argparse
import json
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).parent))
from utils.stream_bootstrap import bootstrap_streams
from utils import importance as _importance
from utils.human_decisions import judge_source_label
from utils.judge_vocab import JUDGE_SKIPPED, upgrade_legacy
from utils.output_handler import get_phase_dir
def load_output(path: str) -> dict:
# An output written before the judge rename says `claude_skipped`.
with open(path, "r", encoding="utf-8") as f:
return upgrade_legacy(json.load(f))
def get_title(item: dict) -> str:
return item.get("title") or item.get("theme") or "Untitled"
def get_importance(item: dict) -> str:
return _importance.normalize(item.get("importance"))
def importance_badge(imp: str) -> str:
return {"HIGH": "HIGH", "MEDIUM": "MED", "LOW": "LOW"}.get(imp, imp)
def format_item_line(item: dict, index: int) -> str:
title = get_title(item)
imp = importance_badge(get_importance(item))
parts = [f" {index}. [{imp}] {title}"]
if item.get("auto_approved"):
parts[0] += f" (auto-approved: {item.get('auto_approval_reason', '')})"
if item.get("edited"):
parts[0] += " (description edited)"
return parts[0]
def display(data: dict, phase: str) -> None:
is_task_suggestions = phase == "apply-task-suggestions"
is_code_fixes = "fix" in phase
if is_task_suggestions:
item_noun_plural = "task suggestions"
elif is_code_fixes:
item_noun_plural = "fixes"
else:
item_noun_plural = "suggestions"
to_apply = data.get("to_apply", [])
needs_human = data.get("needs_human_review", [])
skipped_items = data.get("skipped_items", [])
user_skipped_items = data.get("user_skipped_items", [])
edited_descriptions = data.get("edited_descriptions", [])
summary = data.get("summary", {})
# Build set of edited item IDs for marking in the apply list
edited_ids = set()
for ed in edited_descriptions:
edited_ids.add(ed.get("id", ""))
edited_ids.add(ed.get("title", ""))
# Mark edited items in to_apply
for item in to_apply:
if item.get("title") in edited_ids:
item["edited"] = True
total_groups = summary.get("total_groups", summary.get("total_issues", 0))
valid_count = len(to_apply)
human_count = len(needs_human)
user_skipped_count = len(user_skipped_items)
# Header
if is_task_suggestions:
header = "Task Suggestion Decision Summary"
elif is_code_fixes:
header = "Code Fix Decision Summary"
else:
header = "Suggestion Decision Summary"
print(f"\n{'=' * 60}")
print(f" {header}")
print(f"{'=' * 60}")
print(f" Total groups: {total_groups}")
print()
counter = 1
# Will Apply
auto_count = sum(1 for i in to_apply if i.get("auto_approved"))
label = f"WILL APPLY ({valid_count} {item_noun_plural})"
if auto_count:
label += f" [{auto_count} auto-approved]"
print(f" + {label}")
if to_apply:
for item in to_apply:
print(format_item_line(item, counter))
counter += 1
else:
print(" (none)")
print()
# Needs Human Review
if needs_human:
print(f" ? NEEDS HUMAN REVIEW ({human_count} {item_noun_plural})")
for item in needs_human:
title = get_title(item)
imp = importance_badge(get_importance(item))
reason = item.get("validation_reason", "")
line = f" {counter}. [{imp}] {title}"
if reason:
line += f" -- {reason}"
print(line)
counter += 1
print()
# Invalid / Validation Skipped
invalid_items = [i for i in skipped_items if i.get("status") == "invalid"]
bulk_skipped = [i for i in skipped_items if i.get("status") == "bulk_skipped"]
priority_filtered = [i for i in skipped_items if i.get("status") == "below_priority"]
# The judge's own skips. Bucketed apart from `invalid` because they are the
# opposite kind of outcome: validation said the item was wrong, whereas this
# is a decision the judge agent made on the user's behalf about an item
# validation could not settle — and this command is the one surface whose
# whole job is to say what was decided.
judge_skipped = [i for i in skipped_items if i.get("status") == JUDGE_SKIPPED]
other_skipped = [i for i in skipped_items
if i.get("status") not in ("invalid", "bulk_skipped",
"below_priority", JUDGE_SKIPPED)]
if invalid_items:
print(f" x INVALID ({len(invalid_items)} {item_noun_plural})")
for item in invalid_items:
title = get_title(item)
imp = importance_badge(get_importance(item))
reason = item.get("reason", "")
line = f" {counter}. [{imp}] {title}"
if reason:
line += f" -- {reason}"
print(line)
counter += 1
print()
if bulk_skipped:
print(f" - BULK SKIPPED ({len(bulk_skipped)} {item_noun_plural})")
for item in bulk_skipped:
title = get_title(item)
imp = importance_badge(get_importance(item))
reason = item.get("reason", "")
line = f" {counter}. [{imp}] {title}"
if reason:
line += f" -- {reason}"
print(line)
counter += 1
print()
if judge_skipped:
print(f" ! SKIPPED BY THE JUDGE AGENT ({len(judge_skipped)} {item_noun_plural})")
for item in judge_skipped:
title = get_title(item)
imp = importance_badge(get_importance(item))
reason = item.get("reason", "")
source = item.get("decision_source", "")
line = f" {counter}. [{imp}] {title}"
if source:
line += f" ({judge_source_label(source)})"
if reason:
line += f" -- {reason}"
print(line)
counter += 1
print()
if priority_filtered:
print(f" - BELOW PRIORITY ({len(priority_filtered)} {item_noun_plural})")
for item in priority_filtered:
title = get_title(item)
imp = importance_badge(get_importance(item))
print(f" {counter}. [{imp}] {title}")
counter += 1
print()
if other_skipped:
print(f" - SKIPPED ({len(other_skipped)} {item_noun_plural})")
for item in other_skipped:
title = get_title(item)
imp = importance_badge(get_importance(item))
reason = item.get("reason", "")
line = f" {counter}. [{imp}] {title}"
if reason:
line += f" -- {reason}"
print(line)
counter += 1
print()
# User Skipped
if user_skipped_items:
print(f" ~ USER SKIPPED ({user_skipped_count} {item_noun_plural})")
for item in user_skipped_items:
title = get_title(item)
imp = importance_badge(get_importance(item))
print(f" {counter}. [{imp}] {title}")
counter += 1
print()
# Edited descriptions note
edited_in_apply = [i for i in to_apply if i.get("edited")]
if edited_in_apply:
print(f" * {len(edited_in_apply)} {item_noun_plural} have user-edited descriptions")
print()
print(f"{'=' * 60}")
print()
def resolve_output_file(args: argparse.Namespace) -> str:
if args.output_file:
return args.output_file
if not args.plan_file or not args.phase:
print("ERROR: Provide --output-file or both --plan-file and --phase", file=sys.stderr)
sys.exit(1)
phase_map = {
"apply-suggestions": "apply-suggestions",
"apply-code-fixes": "apply-fixes",
"apply-fixes": "apply-fixes",
"apply-task-suggestions": "apply-task-suggestions",
}
phase_dir_name = phase_map.get(args.phase, args.phase)
phase_dir = get_phase_dir(Path(args.plan_file), phase_dir_name)
return str(phase_dir / "orchestrator_output.json")
def main(argv=None) -> int:
bootstrap_streams()
parser = argparse.ArgumentParser(description="Display decision summary from orchestrator output")
parser.add_argument("--output-file", help="Path to orchestrator_output.json")
parser.add_argument("--plan-file", help="Path to plan file (used with --phase)")
parser.add_argument(
"--phase",
choices=["apply-suggestions", "apply-code-fixes", "apply-task-suggestions"],
help="Phase name: apply-suggestions, apply-code-fixes, or apply-task-suggestions",
)
args = parser.parse_args(argv)
output_path = resolve_output_file(args)
if args.phase:
phase = args.phase
elif "task-suggestions" in output_path:
phase = "apply-task-suggestions"
elif "fix" in output_path:
phase = "apply-code-fixes"
else:
phase = "apply-suggestions"
if not Path(output_path).exists():
print(f"ERROR: Output file not found: {output_path}", file=sys.stderr)
sys.exit(1)
data = load_output(output_path)
display(data, phase)
return 0
if __name__ == "__main__":
sys.exit(main(sys.argv[1:]))
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |