In the DSH ecosystem, plugins are the basic building blocks for agent workflows. Weekly review is a common way for developers to plan their work, but common practice often stops at recording feelings (“It was a busy week and a lot of progress was made”) rather than aligning with action. Such non-closed-loop reviews are essentially just diaries. The weekly-review plugin is designed to address this problem by enforcing strict reconciliation between commitments and outcomes, ensuring that weekly reviews produce actionable follow-ups.

It is a workflow plugin maintained by ChenneyZhuang under the MIT license. Its core value proposition is: a review that changes nothing is a diary. By enforcing strict reconciliation between commitments and outcomes, it ensures that weekly reviews lead to actionable steps.

Core Features

  1. Pull commitments: Retrieve commitments from last week’s review log instead of relying on memory.
  2. Reconcile with evidence: Reconcile commitments against what actually happened, and attach evidence for each status (completed / partially completed / incomplete).
  3. Record evidence: Record completed work as evidence (such as commits, sends, or merges) rather than merely logging feelings.
  4. Defer items: Carry over incomplete items only with a new reason; otherwise discard them explicitly.
  5. Set commitments: Make next week’s commitments few, concrete, and verifiable.

Installation and Enablement

The installation command is:

npx skills add ChenneyZhuang/weekly-review

The plugin version is 0.1.0 and is licensed under MIT.

Typical Usage

  • Review log: Use a single reviews.md file to record once per week (latest entry at the top) so that next week can read it directly.
  • Cross-week comparison: By comparing entries across weeks, you can clearly see drift in work goals.
  • Honest logging: For soft work (such as learning or relationships), when evidence is lacking, honestly recording “no clean evidence” is better than fabricating metrics.

Applicability and Notes

  • Baseline creation: The first review has no historical log to reconcile against, so establish a baseline.
  • Naming evidence: Evidence for soft work is indeed hard to define.
  • Permission check: The plugin runs with the current dsh process permissions; review the source code and license before installation.

Summary

By turning review from emotional logging into an evidence-based closed loop, the plugin helps improve workflow reliability. Full documentation and source code are available on GitHub.