Professional engineering discipline without taking ownership of your workflow.
Meridian gives coding agents one model-independent quality standard, lets projects supply their own customer and repository policy, and applies small model-specific corrections where a model has a repeatable failure mode. It does not route ordinary requests into mandatory planning, review, documentation, or commit pipelines.
Meridian has four layers:
- Professional engineering standard — shared expectations for correctness, evidence, security, maintainability, operability, delivery hygiene, user agency, proportionality, and stopping.
- Project policy — your
CLAUDE.md,AGENTS.md, and repository instructions define customer, compliance, tooling, deployment, rollback, and artifact requirements. - Model adapters — atomic corrections for observable quirks such as comment narration or recursive reviewing. They change steering, never the quality bar or authorization.
- Skills — formal flows are user-owned; narrow engineering disciplines can be selected by the model inside an ordinary task.
The result is deliberately asymmetric: different models may receive different corrective wording, but every model owes the same professional outcome.
These run only when you invoke them. One flow never silently enrolls you in another.
| Skill | Purpose |
|---|---|
brainstorm | Develop an idea into an approved implementation spec |
sketch | Produce a lightweight contract for a small change |
execute | Implement an approved spec or clear scope with proportionate verification |
review | Perform a focused, report-only code review |
document | Capture requested durable engineering knowledge |
auto | Grant elevated autonomy for reversible, in-scope work |
Invoking a flow authorizes its ordinary documented behavior. Separate approval is still required for destructive operations, unrequested external effects, material scope expansion, and unresolved high-cost choices.
The model may reach these inside an ordinary task. They are not workflow gates:
| Skill | Purpose |
|---|---|
commit | Commit and pull request conventions, and the one-clean-commit flow |
debug | Diagnose failures from evidence before fixing them |
research | Verify unstable external behavior against primary sources |
triangulate | Escalate contested or load-bearing claims beyond one source |
delegate | Isolate genuinely independent or context-heavy work |
respond | Evaluate review feedback while preserving user intent |
commit is the exception in this table: it holds conventions and a flow that performs outward actions. The conventions are reachable whenever a commit message or pull request description is being written, including by other skills. Reaching them authorizes nothing: committing, pushing, and opening or editing a pull request sit behind a hard gate and each still require an explicit request.
At session start Meridian injects the professional standard followed by the adapters for the resolved profile. Resolution order is:
MERIDIAN_MODEL_PROFILE- exact model identifier
- registered model-family prefix
- host-local fallback
- baseline standard only
Initial profiles are:
| Profile | Adapters | Mapped by |
|---|---|---|
claude-opus | comment narration, artifact overproduction | claude-opus- prefix |
fable | none | claude-fable- prefix |
sol | review recursion, depth escalation, workflow literalism, agreement seeking | environment override or fallback |
Profiles are named for model families rather than single releases, and families are matched by prefix so a later release in a known family keeps its corrections instead of silently dropping to the baseline.
Matching is a plain prefix over the identifier the host reports. A vendor- or region-qualified identifier, such as the us.anthropic. form used by Bedrock, does not match a bare family prefix; use the environment override or a host-local fallback there.
Only model identifiers observed from host payloads or documented by the host are mapped automatically. A model outside a mapped family receives the baseline rather than a guessed profile.
Set a host-local fallback in <state-base>/meridian/config.json:
{
"fallbackProfile": "sol"
}The state base is host-specific, such as CODEX_HOME for Codex. GitHub Copilot does not currently expose a model identifier in its documented session-start payload, so use the environment override or fallback there.
/plugin marketplace add KodingDev/meridian
/plugin install meridian@meridian
Install Meridian from the Cursor plugin marketplace, or enable third-party skills and install the Claude plugin bundle.
Install the same plugin bundle through the host's plugin support. Meridian ships host-specific hook manifests where their schemas differ. The session hook is read-only context composition; the existing pre-tool commit guard still blocks AI attribution and staging ignored .meridian/ working artifacts where supported.
Use repository instructions for requirements that vary by customer or environment. The included templates/CLAUDE.md provides optional headings for security and compliance, required verification, deployment and rollback, repository artifacts, and allowed tools. Meridian intentionally ships no AWS-, employer-, customer-, or Stella-specific defaults.
pnpm check
node --test test/meridian-hooks.test.mjs
The live evaluation harness is on-demand because it makes paid model calls:
pnpm eval
pnpm eval:view
Built on ideas from Superpowers by Jesse Vincent and informed by Matt Pocock's user-invoked/model-invoked skill split.
MIT