This skill should be used when the user asks to "release the SDK", "prepare a release", "publish a new version", "cut a release", "do a release", or mentions the SDK release checklist or release process. Guides through the full software-agent-sdk release workflow from version bump to PyPI publication, emphasizing human checkpoints.
SkillScore78Maintained
Automatic
◆DevOps & SRE pack
Last verified 9/29/2026
What it changes
Passed 17 of 22 test cases with the skill, and 7 of 22 without it.
Without skill30%
With skill78%
Checks it passed that the model alone missed
Uses the --failed flag when executing gh run rerun.
Specifies passing swebench as the benchmark input parameter.
Specifies querying the repository OpenHands/software-agent-sdk.
and 15 more in the full report
12 fixed6 passed either way4 still failing1 broke
Pass-rate lift
+48 pp
Tokens
+50%
Agent turns
0%
Response time
-49%
Model gemini-3.6-flashMeasured 9/29/2026v118 of 23 cases passed with skill · 22 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 12
Prompt
In `run-eval.yml` dispatches for software-agent-sdk release evaluation, how should the `sdk_ref` parameter be formatted for version 1.25.0?
What was checked, without and with the skill
Checked
Without skill
With skill
Formats the sdk_ref input parameter as v1.25.0 with a leading v.
Passed
Passed
Specifies passing swebench as the benchmark input parameter.
Missed
Passed
case-19 · gemini-3.6-flash · 9/29/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md205 lines2 files
SDK Release Guide
This skill walks through the software-agent-sdk release process step by step.
🚨 CRITICAL: NEVER merge the release PR or create/publish a GitHub
release without the human's explicit approval. Release is the last line
of human defense. Always present the current status and ask for
confirmation before performing any irreversible action.
Phase 1: Trigger the Prepare-Release Workflow
Determine the target version (SemVer X.Y.Z). Then trigger the
prepare-release.yml workflow, which creates a release branch and PR
automatically.
Search, download, and summarize academic papers from arXiv. Use when user says "search arxiv", "download paper", "fetch arxiv", "arxiv search", "get paper pdf", or wants to find and save papers from arXiv to the local paper library.