training-stack-change

Name: training-stack-change
Author: vllm-project/semantic-router

$npx mdskill add vllm-project/semantic-router/training-stack-change

Modifies training workflows and outputs to align with changes in model selection, embedding pipelines, experiments, and evaluation artifacts

Identifies and modifies change surfaces for training stack components under src/training/**.
Deploys tools and APIs within the agent tooling ecosystem to manage and validate changes.
Decides on modifications based on current repo map context and alignment with runtime behavior expectations.
Delivers validated changes through automated reports and gate validations, ensuring all artifacts meet expected standards.

SKILL.md

.github/skills/training-stack-changeView on GitHub ↗

---
name: training-stack-change
category: primary
description: Modifies training-stack workflows, selector or embedding pipelines, evaluation artifacts, or runtime-facing outputs under src/training. Use when changing model selection training, embedding pipelines, evaluation scripts, experiments, or other training outputs that feed runtime behavior.
---

# Training Stack Change

## Trigger

- Change training workflows, scripts, docs, or runtime-facing outputs under `src/training/**`
- Change selector, embedding, evaluation, or experiment artifacts that feed runtime behavior

## Workflow

1. Read change surfaces and the repo map for the current training stack context
2. Modify the training workflow, artifact expectation, or runtime-facing output
3. Run `make agent-report ENV=cpu CHANGED_FILES="..."` to verify surface alignment
4. Run `make agent-ci-gate CHANGED_FILES="..."` to validate all constraints
5. Record any runtime-facing artifact contract mismatches as indexed debt entries

## Gotchas

- `src/training/**` is no longer classifier-only; do not force selector, embedding, eval, or experiment work through classifier language that hides the real contract.
- Training-only edits still change runtime behavior when exported artifacts, defaults, or evaluation gates shift.

## Must Read

- [docs/agent/change-surfaces.md](../../../../docs/agent/change-surfaces.md)
- [docs/agent/repo-map.md](../../../../docs/agent/repo-map.md)

## Standard Commands

- `make agent-report ENV=cpu CHANGED_FILES="..."`
- `make agent-ci-gate CHANGED_FILES="..."`

## Acceptance

- Training workflow changes keep scripts, docs, and artifact expectations aligned
- Runtime-facing artifact contract changes are either updated in code or recorded as tracked debt through the matching indexed debt entry

More from vllm-project/semantic-router

Skill	Description
config-platform-change	Synchronizes config representations across router config, Python CLI schema, and dashboard config UI. Use when adding or changing a config concept that spans those surfaces or addressing config representation debt before Kubernetes-facing translation.
cross-stack-bugfix	Diagnoses and fixes bugs that span multiple layers (runtime, CLI, UI, platform, tests) requiring coordinated changes across surfaces. Use when a bug does not map cleanly to a narrower skill, the fix touches more than one surface, or changes need cross-cutting validation.
dashboard-platform-change	Modifies dashboard frontend or backend surfaces that present, configure, or manage router behavior through the console UI. Use when changing dashboard pages or components, backend handlers, console persistence, auth or session flows, or user-visible routing metadata in the dashboard.
fleet-sim-change	Modifies the fleet simulator package, API service, release wiring, or simulator-owned docs and assets as one maintained subsystem. Use when changing src/fleet-sim, simulator release workflow, or fleet-sim-owned docs and assets under website/.
harness-contract-change	Modifies the repository's agent contract including AGENTS.md, docs index, manifests, validation scripts, and contributor-facing harness wrappers. Use when updating agent documentation, changing repo manifests, editing validation scripts, modifying CI/workflow classification, or updating contributor-facing guides like README.md, CONTRIBUTING.md, or the PR template.
k8s-platform-change	Modifies Kubernetes-facing operator, CRD, deployment-profile, or DSL translation behavior for semantic-router platform integration. Use when changing operator APIs or controllers, deployment stack manifests, profile-owned platform wiring, or router-to-Kubernetes translation layers.
maintainer-issue-pr-management	Manages GitHub issue and pull-request lifecycle including creation, updates, triage labelling, and closeout metadata using canonical templates and repository taxonomy. Use when a maintainer asks to create, update, close, or triage GitHub issues or PRs, or when issue creation requires codebase analysis for scope, labels, or acceptance criteria.
maintainer-release-ops	Maintainer release and milestone operating workflow. Use when a maintainer wants to plan a release, create milestone issues, sync GitHub issue or PR state, generate a daily review brief, or manage stale PRs and backlog routing.
openclaw-vsr-bridge	Install vLLM Semantic Router in agent-safe mode, import supported OpenClaw model providers into canonical VSR config, and rewrite OpenClaw to target VSR.
plugin-end-to-end	Implements end-to-end plugin changes spanning router config, post-decision processing, optional CLI/UI exposure, and E2E test coverage. Use when adding a new plugin type, changing plugin config schema or execution semantics, updating plugin chain behavior, or modifying plugin-exposed metadata across surfaces.