AI Agent Hub
Back to plugins
⚙️

dsh-command-code-review

Workflow Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install JasonFreeLab/dsh-command-code-review

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

Run dsh plugin install JasonFreeLab/dsh-command-code-review in DeepSeek Harness to install this plugin; source repository: https://github.com/JasonFreeLab/dsh-command-code-review

About this plugin

Code review is a critical quality gate, yet manual passes usually cover only one or two dimensions and easily miss security flaws, compliance drift, or the broader history of the code base. dsh-command-code-review packages the /code-review slash command as a single DSH bundle so that a structured, multi-lens review runs entirely inside the agent conversation without context switching to external tooling.

Out of the box the plugin scores findings across five default lenses — style-guide compliance, bugs and correctness, historical context, security, and comment consistency — with an optional performance lens. Findings are deduplicated across lenses, batch-scored for confidence (real vs. false positive) and severity (blocker, major, minor, nit), and anything below a configurable threshold (default 80) is dropped to cut noise. When the review scope touches security-sensitive files or hot paths, the corresponding lens is enabled automatically. Results are written as a structured Markdown report plus a machine-readable JSON sidecar under the doc/ directory, and for pull-request reviews the findings are posted back to the PR thread via the GitHub CLI.

This plugin suits engineers and teams running DeepSeek Harness workflows who want code review to be a repeatable, configurable, and traceable step — particularly in mid-to-large projects where pull requests and local branches coexist and both security and performance reviews must stay in scope.

Use Cases

  • Run a multi-lens review before merging a PR and post findings back
  • Self-check security and compliance on a local branch before committing
  • Enforce team-wide review standards while filtering out low-confidence false positives

Best For

  • Engineers working within DSH agent workflows
  • Mid-size teams standardizing their code review process
  • Developers who prioritize security and performance checks