Pre-cast self-check for the main agent. Do I actually need clones? Which of the 9 modes? Am I within my subscription's usage/rate limit? Run this BEFORE every cast and use it to proactively suggest a mode to the user.

2026-06-092

manta-orchestrate

tr00x/Manta

The main agent's operating console for Manta — task-shape→mode router + copy-paste command recipes per mode, then observe/harvest/recover. Load BEFORE any manta cast / /manta:* / manta_* tool. Be proactive — when a task fits a cast shape, offer it.

2026-06-092

manta-pair-reviewer

tr00x/Manta

Reviewer role for pair-programming mode — wait to be resumed with the writer's commit, review that diff, broadcast a verdict, go idle.

2026-06-032

manta-pair-writer

tr00x/Manta

Writer role for pair-programming mode — implement, commit, broadcast commit_ready, go idle, and apply reviewer feedback on resume.

2026-06-032

manta-as-clone

tr00x/Manta

Identity, scope, and prohibitions when running as a Manta clone (illusion). Loads first thing on clone startup.

2026-06-022

manta-coordinate

tr00x/Manta

File locks, broadcast etiquette, work-claim board. How a clone shares the bus without stepping on siblings.

2026-06-022

name	manta-merge-review
description	Guide for the main agent to interpret merge-review output and promote a winner from forking-realities casts
audience	main
version	0.0.1
related	["manta-cast-decide","manta-as-clone"]

manta-merge-review

Purpose

After a forking-realities cast completes and all clones are DEAD, the CLI automatically generates docs/merge-reviews/<castId>.md with a scored ranking of all candidates. This skill describes how the main agent reads the review, interprets scores, and promotes a winner.

Allowed

Read merge-review documents from docs/merge-reviews/<castId>.md.
Override the scoring engine's recommendation when domain knowledge suggests the lower-scored candidate is better.
Promote any clone in the roster, not just the top-ranked one.
Inspect individual worktree branches before promoting to validate solution quality.
Re-run merge-review if metrics look stale (e.g. after manual fixes in a worktree).

The merge-review document contains:

Verdict — auto_merge_eligible | manual_review_required | no_candidates_passed_gate | dominance_inversion_flagged.
Score table — per-candidate normalized scores on 6 axes (coverage, diff, complexity, typeCheck, lint, perf) plus composite score.
Disqualified section — candidates whose test suite failed.
Tie-break explanation — axis priority → Pareto dominance → self-certainty → defer to main.
Weight adjustments — agentic rubric pre-pass adjustments based on project config.
Proposed merge command — the exact git merge command.

Promote via /manta promote <castId>/<cloneId>. This merges the winner's branch (--no-ff), graveyards loser worktrees, removes the winner worktree, and emits a promote event.

Forbidden

Auto-promoting without reading the review.
Ignoring dominance_inversion_flagged verdicts — they indicate weight mis-calibration.
Promoting from a cast that has no merge-review event (the command will reject this).

Examples

A typical merge-review flow:

Cast completes: manta cast forking-realities --clones 3 --task "implement feature X".
All clones die. CLI auto-generates docs/merge-reviews/cast-123456.md.
Read the review: clone B scores 0.82, clone A scores 0.79, clone C disqualified (test failure).
Verify: git diff main...manta/cast-123456/B — confirm the diff is sane.
Promote: /manta promote cast-123456/B.
Result: B's branch merged, A's worktree moved to graveyard, C's worktree moved to graveyard.