AI Agent Hub
Back to plugins
🧰

dsh-onfail

Web Tools Updated 2026.08.28

Run the following command in DeepSeek Harness:

dsh plugin install ZZZjf13960/dsh-onfail

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install ZZZjf13960/dsh-onfail in the DeepSeek Harness terminal to install the plugin; the source repository is available at https://github.com/ZZZjf13960/dsh-onfail.

About this plugin

The most flow-breaking moment in a CI failure is leaving your DSH session to hunt for logs in a browser tab. dsh-onfail pulls that moment into your sidebar: it polls the Actions runs of the workspace you are in (auto-detected origin and branch, following the open PR head when one exists) and surfaces only new failed checks as compact cards showing the job name, SHA, and a trimmed log excerpt.

Clicking a card does not spawn a new session and does not steal focus. It injects the repository, job, SHA, failing step, and a short log snippet directly into the queue of the session you are already in, so you can start fixing with full context in front of you. Each job + SHA is deduplicated persistently, so the same completed failure will not restack, while a rerun that produces a new job id is treated as a fresh problem and shown again.

It is built for DSH users who live in GitHub Actions day to day: no webhooks, no sidecar service, no context switch. Wire up a token (or just reuse gh auth login) and the plugin sits quietly in the sidebar footer, ready the moment a red check appears.

Use Cases

  • A CI failure card appears in the sidebar; one click continues your active session with job, SHA, and log context
  • Polls the current branch Actions runs and surfaces only new failures without restacking historical ones
  • A rerun that generates a new job id re-triggers the card, while the same completed SHA stays deduplicated

Best For

  • DSH users who rely on GitHub Actions in their daily workflow
  • Developers who want to fix CI failures without leaving the current session
  • Engineers using DSH Web profiles for day-to-day build and repair cycles