# GitHub Action: run skill-up Agent Skill evals in CI, across multiple engines.
#
# This metadata file lives at the repo ROOT and the repo contains a SINGLE
# action — both are required for GitHub Marketplace listing/discoverability.
# `branding` (icon + color) is what renders on the Marketplace card.
name: 'Agent Skill Eval (skill-up)'
description: 'Run skill-up behavioral evals on your Agent Skills in CI — across claude_code, codex, qodercli, and qwen_code engines.'
author: 'skill-up'

branding:
  icon: 'check-circle'
  color: 'purple'

inputs:
  engine:
    description: 'Agent engine: claude_code / codex / qodercli / qwen_code. Leave empty to let the repo''s eval.yaml declare it.'
    required: false
    default: ''
  model:
    description: 'Passed to `skill-up run --model` when non-empty; otherwise the eval config / CLI default is used.'
    required: false
    default: ''
  provider:
    description: 'Resolves base-url from a built-in table (public: dashscope). Ignored when base-url is set.'
    required: false
    default: ''
  api-key:
    description: 'Model credential. engine non-empty -> routed to ANTHROPIC_API_KEY / OPENAI_API_KEY / QODER_PERSONAL_ACCESS_TOKEN; engine empty -> passed via `skill-up --api-key`. Always source this from a GitHub secret.'
    required: false
    default: ''
  base-url:
    description: 'Model API endpoint (public). Defaults to the agent CLI''s own public endpoint when empty.'
    required: false
    default: ''
  open-sandbox-api-key:
    description: 'OPENSANDBOX_API_KEY; only needed when evals use the opensandbox environment.'
    required: false
    default: ''
  skill-target:
    description: 'Skill directory or eval config path, resolved against the checked-out repo root.'
    required: false
    default: '.'
  skill-up-version:
    description: 'Fallback skill-up version for custom images without skill-up. The official image preinstalls a pinned version and ignores this input.'
    required: false
    default: '0.7.0'
  skill-up-command:
    description: 'The skill-up command itself; must start with `skill-up`.'
    required: false
    default: 'skill-up run'
  parallelism:
    description: 'Passed to `--parallelism` when non-empty (1-256).'
    required: false
    default: ''
  agent-install-command:
    description: 'Optional custom install command, for a custom image missing a target engine CLI.'
    required: false
    default: ''

outputs:
  exit-code:
    description: 'skill-up exit code (0 = all evals passed).'
  report-dir:
    description: 'Directory with the JSON / JUnit / HTML reports, relative to the workspace (e.g. skill-up-workspace), so host steps can consume it.'

# Docker container action: the prebuilt image ships the three engines + skill-up
# + the action code, so the caller just `uses:` it. The image is published to
# this repo's own org namespace by .github/workflows/runner-image.yml;
# bump the tag together with releases (pin by @sha256 digest for reproducibility).
# NOTE: Docker container actions run on Linux runners only.
runs:
  using: docker
  # v0.1.0; update the digest deliberately when publishing a verified image.
  image: docker://ghcr.io/alibaba/skill-up-runner@sha256:f90aa1a320dadd14f354c12013dbbf78df262ece0e4f82683b8a70140a65bbcf
  args:
    - --engine
    - ${{ inputs.engine }}
    - --model
    - ${{ inputs.model }}
    - --provider
    - ${{ inputs.provider }}
    - --api-key
    - ${{ inputs.api-key }}
    - --base-url
    - ${{ inputs.base-url }}
    - --open-sandbox-api-key
    - ${{ inputs.open-sandbox-api-key }}
    - --skill-target
    - ${{ inputs.skill-target }}
    - --skill-up-version
    - ${{ inputs.skill-up-version }}
    - --skill-up-command
    - ${{ inputs.skill-up-command }}
    - --parallelism
    - ${{ inputs.parallelism }}
    - --agent-install-command
    - ${{ inputs.agent-install-command }}