Capability contracts
Every meaningful behaviour has a machine-checkable input/output contract with real examples — executable specs, not prose.
A repo standard for governable vibe coding.
An agent-independent repository standard that keeps AI-generated software understandable, testable and maintainable as it grows — it brings memory, decisions, boundaries and evidence into your repo, so AI-assisted projects can grow without losing control.
Spec-driven tools tell an agent what to build. Trellis preserves why it was built that way, how each piece is verified, how much it may grow, and when it must be reconsidered.
Every meaningful behaviour has a machine-checkable input/output contract with real examples — executable specs, not prose.
Source → Decision → Capability. Every claim traces back to the source and the ADR that justify it, validated in CI.
Technical and semantic limits — files, lines, dependencies, responsibilities — enforced, so the codebase can’t quietly sprawl.
Declare which models may author code. A degraded fallback model is caught and blocked before it can merge.
A companion tool that draws your repository’s evidence graph as one self-contained page. One command, no server, opens offline.
The rules you adopt in your repo — contracts, decisions, budgets, evidence. Runs in CI.
An optional tool that turns that same evidence into a graph you can explore. It only reads; it never changes your code.
any node to trace its links — everything else dims.
to inspect a node: its claim, source and the decisions that govern it.
one self-contained HTML file — no server, no network, no telemetry.
Paste a single line into your agent — no files, no wizard:
Fetch https://raw.githubusercontent.com/e-onux/trellis/main/INSTALL.md and follow it.
Drop the bootstrap manifesto in and let the agent scaffold:
curl -o TRELLIS.md https://raw.githubusercontent.com/e-onux/trellis/main/TRELLIS.md
Deterministic scaffold and audit for pipelines:
npx @sidrelabs/trellis init --preset standard