Fission Laravel Package

Technical Evaluation

Positioning. joshcirre/fission is not a library you depend on — it is a fork-and-own starter kit. That single fact drives every other conclusion in this assessment: there is no composer require upgrade path, no semantic-version contract, and no upstream patch channel. Integration cost is front-loaded; maintenance cost is permanent and ours.

Architecture fit

• Strong fit only if the target product is a server-driven, session-based Laravel UI on the TALL stack. The kit is opinionated around Livewire 4 + Flux 2 + Tailwind 4, with Pest 4 for tests and Larastan/Rector/Pint for quality gates.
• Team-awareness is baked in as a first-class convention (current_team_id on the user, team-scoped URLs such as /{team}/playground, team_id on owned records, authorization via membership + TeamPolicy). This is an asset for multi-tenant B2B products and dead weight plus surface area for single-tenant or API-first products.
• The agents/AGENTS.md layer is the genuine differentiator and the main architectural coupling. It ships Laravel Boost v2 (MCP server), version-pinned ecosystem context, and seven activation-gated skills (laravel-best-practices, livewire-development, fluxui-development, pest-testing, tailwindcss-development, blaze-optimize, instruckt). Effectively the kit ships an opinionated engineering operating model, not just scaffolding — conventions, Artisan-first workflow, "every change must be programmatically tested," Pint-before-finalize, search-docs before edits.
• It assumes a Bun-based frontend pipeline (bun run build, composer run dev) and the Boost MCP toolchain (database-query, database-schema, get-absolute-url, browser-logs). Teams on npm-only or without MCP-capable editors lose part of the value proposition.

Integration feasibility

• Greenfield is straightforward: clone/template → composer install → bun install → env/DB → migrate/seed → build. A competent Laravel developer reaches a running app in under a day.
• Retrofit into an existing application is not recommended. The value is in conventions applied from commit zero; grafting team-scoped routing and Flux components onto a mature codebase yields the churn without the coherence.
• Hard prerequisites: PHP 8.4 runtime and CI, Laravel 13, Livewire 4, Flux 2 (free) plus Flux Pro 2 (commercial, per-seat), Tailwind 4, Node + Bun, and a database supported by Laravel 13. Anything less than a current-gen platform makes this a platform-upgrade project disguised as a starter kit.
• The dependency set is uniformly bleeding-edge (PHP 8.4, Laravel 13, Livewire 4, Flux 2, Pest 4, Tailwind 4, Larastan 3, Rector 2, PHPUnit 12). Feasibility is therefore gated less by this package than by our appetite for running near the tip of every ecosystem simultaneously.

Technical risk

Key questions

Integration Approach

Stack fit. Adopt as a reference architecture and template, not as a dependency. Best-fit profile: new greenfield Laravel product, server-rendered UI, small-to-mid team, existing Laravel fluency, multi-tenant ambition, and a team already using or willing to use AI-assisted development with MCP.

Migration path. There is no migration — there is adoption. Sequence it as a scaffold decision, not a code change:

1. Clone into a sandbox repo; run the documented setup; build and test clean.
2. Replace branding, auth flows, and any team scaffolding we will not ship.
3. Establish the baseline: Pint, Larastan, Rector, Pest, and CI green on day one.
4. Only then create the real repository, with a first commit that is a reviewed, pruned snapshot — and a README stating the upstream commit it derives from.

Compatibility. Pin everything: composer.lock, bun.lock, and the exact PHP minor in CI and in the deploy image. Treat AGENTS.md's version manifest as a contract to re-verify at each dependency bump. Verify Flux free-vs-Pro component availability early — a Pro-only component discovered in week three is a budget conversation, not a technical one.

Sequencing

Do not parallelize phases 1 and 2. The purpose of phase 1 is to discover whether the opinionated conventions actually hold under our real requirements before we amplify them across teams.

Operational Impact

Maintenance. Ownership transfers fully on clone. Expect recurring, non-optional work: upstream diff review each Laravel/Livewire minor, AGENTS.md version-pin refresh, Pint/Larastan rule reconciliation, and Flux/Tailwind major migrations. Realistically 2–5 engineer-days per quarter in steady state, spiking to 1–2 weeks at each ecosystem major. Assign this to a named owner with calendar time; unowned starter kits rot into unupgradeable snowflakes within two cycles.

Support. Community-only for the kit itself, with commercial support limited to Flux Pro. No SLA, no security advisory feed, no CVE notification path for the scaffold. We must subscribe to Laravel, Livewire, Flux, and Tailwind advisories directly and treat the kit's contents as our code for patch purposes. Boost/MCP and the skills are tooling, not supported infrastructure — breakage there is a developer-experience incident, not an outage.

Scaling. The starter kit itself is not a scaling determinant; it is a per-request PHP app whose scaling profile is set by our architecture choices. Two Livewire-specific considerations matter: server-side state means session affinity or shared session storage behind a load balancer, and Livewire round-trips generate far more requests per user session than a classic page-based app, which shifts load from bandwidth to PHP workers and requires queue-backed offloading for anything expensive. Flux is client-side and largely inert. None of this is disqualifying — it just must be designed in before traffic, not after.

Failure modes

• Template stall: dependency majors arrive, nobody owns the diff, the app freezes on Laravel 13 while the ecosystem moves. Most likely and most expensive failure.
• Convention cargo-cult: agents faithfully apply AGENTS.md rules that no longer match installed versions, producing plausible but wrong code. Ground truth must remain the test suite, not the guideline file.
• Feature creep from the scaffold: team routing, playground routes, and demo components ship to production because removal felt like scope. Prune aggressively at phase 1.
• License lapse: Flux Pro renewal missed; UI degrades or compliance is questioned mid-quarter.
• Tooling exclusivity: Boost MCP becomes a soft requirement, narrowing the pool of contributors whose editors support it.
• Security-blind scaffold: no advisory channel means framework patches arrive via our own monitoring or not at all.

Ramp-up. A Laravel-fluent engineer is productive in 2–3 days and confident in 1–2 weeks. The real learning edges are Livewire 4's component model and reactivity, Flux component composition and theming, and the agent-assisted workflow itself. The AGENTS.md skills shorten onboarding meaningfully for Livewire, Flux, Tailwind, and Pest — but they also mask gaps: developers can produce idiomatic-looking code without understanding the request lifecycle, which surfaces later as debugging debt. Pair the skills with deliberate review of Livewire state and query behavior in the first month, and treat the "every change must be programmatically tested" rule as the actual onboarding curriculum. Budget roughly one engineer-week of senior pairing per two new joiners, and one day of enablement for teams already on the TALL stack.

TPM recommendation. Conditionally adopt, scoped to greenfield TALL-stack products, contingent on the phase 0 gates (PHP 8.4/Laravel 13 in CI and prod, Flux Pro seats confirmed, team-feature scope decided, named maintenance owner). Do not adopt for API-first, SPA, or existing-app retrofit work. Re-evaluate at the end of the pilot against our current starter baseline; with 213 stars and a low opportunity score, the case rests on the encoded conventions and agent workflow, not on the code volume.