Skip to content

feat(bosses): a distinct fight for each of the other seven bosses - #40

Merged
vriesd merged 8 commits into
mainfrom
feat/boss-movesets
Nov 13, 2026
Merged

vriesd merged 8 commits into
mainfrom
feat/boss-movesets

Conversation

@vriesd

@vriesd vriesd commented Nov 13, 2026

Copy link
Copy Markdown
Contributor

Boss fire patterns had too little variation: after the boss engine landed, only Bastion Bulwark had a real moveset and the other seven bosses still fired one attack per phase on repeat. This authors a fight for each of them, config only, each in its own level file.

  • Pyre Herald (sector 3): stays the gentle first boss. Two slow attacks per phase, fewer shots per second than before, no escalation.
  • Marshal Vectra (4): introduces anchors, and a crossfire that leaves a safe pocket under the hull.
  • Reef Stalker (5): three phases and the first aimed burst.
  • Choir Regent (7): three movements of fans on a beat, each leaving a different lane open.
  • Umbral Ark (8): every phase opens by launching escorts in front of its guns; the last adds a wall whose gap moves.
  • Hive Maelstrom (9): one turning ring, then counter-rotating rings, then rings that chase you.
  • Omega Null (10): four phases that reprise the campaign, with the style switch at 50%.

The guns get a role without any being required: escorts reward Lance, safe close-range openings reward Scatter, and adds pull Seekers off the boss.

Hull was retuned per boss so fights stay near par instead of getting longer. Mean kill time with the scripted bot (four guns each, one run at 60 fps, no deaths), against par: sector 3 40.5 s / 45, 4 47.3 / 50, 5 50.3 / 52, 6 51.3 / 60, 7 63.4 / 58, 8 63.1 / 62, 9 57.9 / 65, 10 68.1 / 70. Lance is the slowest gun against bosses and runs past par on sectors 7 and 10 (77 s and 83 s). Run-to-run noise is about ten seconds, and this has not had a human playtest.

Proof: typecheck, lint, format, knip, unit tests (level validation covers every boss's phases and par), build and bundle check pass; the full browser suite passed locally on desktop and mobile except one load-sensitive font-timing smoke test that these files do not touch (95 passed, 5 skipped by design). I looked at screenshots of each boss on desktop and phone portrait.

Reviewed with sol-loop: 2 rounds with gpt-6.1-sol. Sol verdict: approved

Created with Claude Fable 5.2 in Claude Code (orchestration, integration); fights authored by Claude Opus 5.5.

🤖 Generated with Claude Code

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 13, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-13T17:00:48.035182Z c4a2075 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

vriesdcrv and others added 8 commits November 13, 2026 18:09
The sector 7 boss fired one fan pattern per phase on repeat and leaned on
hull, so its fight felt long and samey. It now has three phases, each a
fixed rhythm of volleys that leaves a different lane open:

- Hymn: two interleaving fans, then a fan aimed at the player.
- Antiphon (66%): outward loft fans keep the lane under the boss open,
  so flying in close (where Scatter is strongest) is safe; a nave fan
  then closes that lane, and a cantor's aimed burst follows the player.
- Crescendo (33%): a rotary, the hymn and the lofts with a short burst,
  over a faster patrol.

Hull drops from 1100 to 1040 so the measurement bot's mean kill lands
on the 58 s par (57.1 s over two runs of all four guns).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The sector 8 carrier fired one pattern per phase and its summons were
an afterthought. It now has three phases, and each opens with a launch
that puts escorts between the player and the hull:

- Hangar: a fighter pair, then two crossing broadsides that leave the
  flanks open.
- Convoy (66%): a gunship rides ahead, flak follows the player, and the
  rotary turns.
- Narrows (33%): a wall of shots whose gap moves each time, with a
  dodger pair launched into the corridor.

Escorts in the firing line reward Lance (it pierces them on the way to
the hull) and pull Seekers off the boss, so that pick becomes a
trade-off; the standard gun stays fully viable. Hull rises from 1120 to
1320 so the bot's mean kill lands on the 62 s par (62.6 s over two runs
of all four guns), instead of under it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…hases

Hive Maelstrom fired one fan, then one spiral, on repeat. Now it teaches
its spin before it hunts with it: a single ring fired twice so the turn is
easy to read, then two rings turning against each other from a fast
patrol, then at one third it slides after the player with its fastest
rings. Swarm pairs (splitters in the last phase) launch beside the hull on
every other round, so Lance clears them on the way to the boss and Seekers
pay for chasing them without being starved. Hull and par are unchanged:
the scripted bot kills it in 43-76 s (mean 59 s) against a 65 s par.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The finale fired one fan, then one spiral, and soaked up the most hull in
the game. It now reprises what earlier bosses taught, one lesson per
phase: interleaving fans with an aimed burst, a shielded wall with a
gunship guard that Lance cuts through, counter-rotating rings and a low
flank beam that rewards flying close with Scatter, and a last phase that
hunts the player with a beam inside sparse rings, an aimed burst and a
shielded wall in turn. It still switches from the pressure hull to the
maelstrom hull at half health.

Hull drops from 1880 to 1300 because the fight's time now goes into
reading attacks: the scripted bot kills it in 56-72 s (mean 65 s) against
the unchanged 70 s par, taking 3-5 hull hits and surviving.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first boss fired one pattern per phase on repeat. It now alternates
two slow, widely spaced attacks per phase, so a new player sees more than
one volley to read without the fight getting harder: twin ember fans and
a fan that leans with its patrol, then a turning spiral and the game's
first aimed fan. It still never escalates and needs no gun pick; hull is
retuned (275 -> 335) so the measured kill lands under its 45 s par.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Marshal Vectra fired one fan, then one fast spiral, on repeat. Its fight
now teaches anchors, one new idea for sector 4: two line fans to step out
of, then a crossfire that leaves a pocket under the hull, so closing in
is the reward (Scatter shines there; every gun works). At half hull it
steps forward to a centre post for the crossfire and adds a spiral to its
march. Hull drops 585 -> 505 so the measured kill sits under its 50 s par.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Reef Stalker chased with one aimed fan, then one faster double fan. It is
now a three-phase hunt that debuts the aimed burst: shots re-aimed one by
one, so the player has to keep moving. In the middle third it breaks off
the chase to fan across the screen from each edge, where the way out is
under its own hull; the last third chases faster with longer bursts and a
wide shatter fan. Hull 615 -> 625 keeps the measured kill under 52 s par.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@vriesd
vriesd force-pushed the feat/boss-movesets branch from c4a2075 to 4967fde Compare November 13, 2026 17:09
@vriesd
vriesd merged commit bbe4302 into main Nov 13, 2026
10 checks passed
@vriesd
vriesd deleted the feat/boss-movesets branch November 13, 2026 17:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants