AI Agent Hub
Back to plugins
🧩

dsh-stack

admin-security Updated 2026.09.10

Run the following command in DeepSeek Harness:

dsh plugin install Grove-ovo/dsh-stack

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

Run dsh plugin install Grove-ovo/dsh-stack in the DeepSeek Harness terminal to install the plugin; the source repository is at https://github.com/Grove-ovo/dsh-stack, and it activates automatically into the default profile cordis bundle layer after installation.

About this plugin

In a stacked PR workflow, a single rebase at the base layer rewrites every SHA above it, instantly invalidating CI results and review approvals. Engineers see a Sync Success message and reflexively hit merge, yet the new HEAD was never re-validated. dsh-stack is designed as a DeepSeek Harness safety-rail plugin to close exactly that blind spot.

It enforces eight guard assertions across the full lifecycle: environment and auth validation, clean-worktree pre-flight, dependency topology integrity, per-layer test re-runs after every cascade rebase with SHA-bound validation records, landing eligibility checks, MERGED-status polling confirmation, and zero-dependent branch cleanup. /stack-land refuses to merge any layer that lacks a passing record for its current HEAD, while branch deletion is SHA-conditional and uses --force-with-lease so a branch that moved between the check and the delete is retained rather than destroyed. Every command supports --json output for stable consumption by AI agents and CI/CD pipelines.

If you already manage multi-layer dependency chains with gh-stack, or want tested to become provably tested per commit, dsh-stack layers defense in depth on top of GitHub branch-protection rules, making every merge auditable and every cleanup safe and reversible.

Use Cases

  • Re-run tests on every rebased layer and produce SHA-fingerprinted validation records
  • Block /stack-land on any layer that lacks a passing record for its current HEAD
  • Poll GitHub for MERGED state, then conditionally prune zero-dependent branches

Best For

  • GitHub developers managing multi-layer dependency chains with gh-stack
  • Solo maintainers who want per-commit test outcomes to be provable
  • Engineering teams wiring PR merge workflows into AI agents or CI/CD pipelines