| 1 | { |
| 2 | "skill_name": "release", |
| 3 | "evals": [ |
| 4 | { |
| 5 | "id": 1, |
| 6 | "prompt": "Cut release v0.4.3 for onde-cli. Bump all version pins, follow the repo's release branch flow, tag, and push.", |
| 7 | "expected_output": "Agent bumps Cargo.toml, pubspec.yaml, server.json (3 spots), commits signed with Co-Authored-By trailer, creates release/v0.4.3 from development, merges --no-ff into development then main, tags the main merge commit, pushes development+main+tag, and watches the release pipeline.", |
| 8 | "files": [] |
| 9 | }, |
| 10 | { |
| 11 | "id": 2, |
| 12 | "prompt": "We fast-forwarded main to the version bump commit by mistake and tagged it. development is now behind main. Fix the repo state without rewriting pushed history.", |
| 13 | "expected_output": "Agent recognizes the git skill's rule (tag belongs on a merge commit; development must carry the release state) and back-merges main into development with --no-ff instead of rewriting main or deleting the pushed tag.", |
| 14 | "files": [] |
| 15 | }, |
| 16 | { |
| 17 | "id": 3, |
| 18 | "prompt": "The v0.4.2 tag push failed the crates.io verify step because pubspec.yaml still said 0.4.1. Nothing was published. Walk me through fixing it and re-releasing.", |
| 19 | "expected_output": "Agent fixes the pubspec version on a new release branch or hotfix, re-runs the release flow, deletes and recreates the tag only if crates.io never published, and explains why crates.io versions are immutable so a published 0.4.2 would require 0.4.3 instead.", |
| 20 | "files": [] |
| 21 | }, |
| 22 | { |
| 23 | "id": 4, |
| 24 | "prompt": "Update the README install instructions for the npm wrapper and fix the NuGet README badge that points at the wrong package id.", |
| 25 | "expected_output": "Should NOT deeply engage the release skill beyond reading it; this is distribution/README work. Agent edits npm/onde-cli/README.md and nuget/onde-cli/README.md without cutting a release, bumping versions, or tagging.", |
| 26 | "files": [] |
| 27 | }, |
| 28 | { |
| 29 | "id": 5, |
| 30 | "prompt": "Merge the feature/lora-scheduler branch into development.", |
| 31 | "expected_output": "Should NOT trigger the release skill; this is a normal feature merge governed by the git skill (--no-ff into development, no tag).", |
| 32 | "files": [] |
| 33 | } |
| 34 | ] |
| 35 | } |