Three guides recommend the same thing. The one who finished the game disagrees.
Helm Splitter is the near-universal early pick: RPG Division builds around it, Arekkz calls it "really safe", and IGN recommends investing heavily in it for groups and bosses alike. All three are betting on the same thing, which is that being above an enemy is a good place to be. Skill Up, reviewing after completing the game, says the opposite in as many words — enemies "track you in the air really effectively", their overheads reach you up there, and it is "much safer to stay on the ground all of the time, especially against bosses". Read carefully, Arekkz and Skill Up may not conflict at all: Arekkz scopes his safety claim to normal enemies and groups, Skill Up scopes his warning to bosses, and both can be true at once. The real collision is RPG Division against Skill Up, and on that one RPG Division states his own expiry condition on camera — that later enemies may invalidate the whole combination.
What is actually established about each skill
Ordered by evidence rather than by strength. Four records have been watched working in somebody’s footage, which is the highest tier anything in this game reaches; several more have had their in-game panel photographed, which settles the wording without settling what it does in a fight; and the rest are names on a tree with no published effect at all. Filter by Emma or Koo, and note the two rows carrying a dispute marker — they are the ones the argument above is about, and they are also two of the best-evidenced records here. Strong evidence that a skill exists and works is not evidence that building around it is wise.
15 skill records · 4 watched working on camera
Ordered by how well established each record is, because that is the only axis this game supports. No source publishes a damage figure for any node, so a tier list would be somebody’s opinion wearing a table’s clothes.
Creator-tested4 records
Somebody was filmed using it and the effect is visible in the footage. This is the strongest evidence anything in this game has.
- Helm SplitterAirborne area damage and stagger
- TauntPulls enemies off Emma
- PounceOpener carrying an assassination chance
- Full BloomDamage window on Koo, not a spendable art
In-game tooltip captured5 records
The skill’s own in-game panel has been photographed, so the description is the game’s rather than a transcription of it. What the numbers mean in a fight is still untested.
- ShakuyakuStagger pressure, and the gate on Helm Splitter
- IaidoCrowd clear out of a paused combo
- Higanbana: Bloom, Seeds and RootsKoo’s three flower Bloom Arts, equipped into his art slots
- Tracking ShotBranch root: the whole ranged half of Emma’s tree
- Entanglement OverdriveTime-slow burst
Name confirmed; tooltip pending5 records
The name is solid — it appears on the tree or in a guide — and nothing beyond the name is established. Several of these have no published effect at all.
- RollDisengage
- HairBranch root: the one three creators tell you to open first
- IncantationBranch root: Koo’s mod branch
- GuardKoo node named as an early priority
- SnatchAutomatic support; effect not published
Transcription-sensitive1 record
Transcription-sensitive: the name itself comes from machine captions or a single forum answer and may be wrong.
- Deft PawsPassive resupply
The two rows marked safety disputed are the ones every recommendation in this game is built on — and the only source who had finished the game when he spoke says the style they enable is “much safer” abandoned against bosses. All four positions, side by side →
The one claim that would settle it, and nobody has filmed it
Underneath the whole dispute sits a single mechanical question: can Emma block or parry while airborne? Skill Up says she cannot, and he is the only person who says it either way. If he is right, every aerial build carries a defensive cost none of its authors priced in, and the case for staying on the ground against bosses is close to settled. If he is wrong, most of his boss argument falls over. It is a ten-second clip. Nobody has published it, and until somebody does, a site telling you confidently which skills are best is telling you something it cannot know.
Attributes: the one explicit recommendation is conditional, and says so
The attribute question gets asked alongside the skill one and has even less published behind it. Exactly one creator makes a specific call — RPG Division puts points into Vitality — and he immediately scopes it to hard difficulty, saying that on normal or easy you should balance attributes or spend them elsewhere and explicitly handing the decision back to the player. Copying the spread without copying the difficulty is copying the wrong half. Nobody has published what any attribute does per point, what the thresholds are, or whether any of them gate equipment. A "best attributes" table anywhere else is a table with nothing under it.
Where the tree itself is disputed
One structural correction worth carrying before you spend anything: IGN describes the route to Helm Splitter differently from the two creators who filmed themselves taking it. Both of them put it on the hair branch, and they recorded independently on the same day without either being able to copy the other. Where a written guide and two separate captures disagree about the shape of a menu, the captures are the better evidence — but the disagreement is published here rather than resolved, and the skill-trees page carries the full version along with what committing a branch costs.