Agent Skills: Cherry Pick

>

UncategorizedID: opendatahub-io/ai-helpers/vllm-backport-cherry-pick

Repository

opendatahub-ioLicense: Apache-2.0
3674

Install this agent skill to your local

pnpm dlx add-skill https://github.com/opendatahub-io/ai-helpers/tree/HEAD/helpers/skills/vllm-backport-cherry-pick

Skill Files

Browse the full folder contents for vllm-backport-cherry-pick.

Download Skill

Loading file tree…

helpers/skills/vllm-backport-cherry-pick/SKILL.md

Skill Metadata

Name
vllm-backport-cherry-pick
Description
>

Cherry Pick

Selects eligible candidates from ranked.json, attempts cherry-pick on each, and creates a draft PR if any succeed.

Candidate Selection

A PR is eligible if ALL of:

  • backport_ease == "ai-fixable"
  • score >= 50
  • already_backported == false
  • verdict is must_backport or likely_relevant

Usage

bash scripts/cherry-pick.sh \
  --input artifacts/backport-triage/ranked.json \
  --downstream /path/to/downstream-repo \
  --branch rhai/0.13.0 \
  --jira-url "https://redhat.atlassian.net/browse/..." \
  --report-url "https://github.com/..." \
  --output artifacts/backport-triage/cherry-pick-result.json

Output

cherry-pick-result.json:

{
  "status": "created|skipped",
  "pr_url": "https://...",
  "succeeded": 3,
  "conflicts": 1,
  "results": [{"number": 12345, "title": "...", "score": 85, "status": "success|conflict"}]
}

Agent Follow-up (Required)

After this skill runs, the agent MUST:

  1. Semantic validation — review the cherry-picked diff, check imports reference modules that exist at the target tag, check for calls to post-release functions
  2. If issues found, add a comment on the draft PR
  3. Update the Jira ticket with the PR link
  4. For conflict candidates with score >= 70, add label ai-autofix-candidate