Code Review Workflow

The aitasks review system separates code quality review from implementation. The philosophy: first build something that works, then review it for quality improvements. This separation of concerns makes more efficient use of the LLM context — implementation and review are distinct cognitive tasks that benefit from dedicated focus.

The Review Cycle

After completing an implementation task with /aitask-pick, run a targeted review:

1. Run /aitask-review — Select the files or commits to review, and choose which review guides to apply. The skill auto-detects the most relevant guides based on the languages and frameworks in your code
2. Review findings — Examine the results grouped by guide and severity. Select which findings to act on
3. Create tasks — Turn selected findings into tasks (single task, grouped by guide, or one per finding)
4. Implement fixes — Run /aitask-pick on the review task to fix the issues

This cycle can be repeated with different review guides to cover multiple quality dimensions (security, performance, style, etc.).

Walkthrough: Reviewing a Shell Script Module

You’ve just finished adding several new helper scripts to .aitask-scripts/. Before moving on, you want to check them against shell scripting best practices.

1. Launch the skill

/aitask-review

Select “Specific paths” when prompted and enter .aitask-scripts/ as the target.

2. Select review guides

The auto-detection system analyzes the target files, finds .sh extensions and bash shebangs, and ranks shell-related guides first. You’re presented with a multi-select list:

• Shell Scripting (score: 7) — top match
• Error Handling (universal)
• Code Conventions (universal)
• Security (universal)
• …

Select “Shell Scripting” and “Error Handling” — two guides keeps the review focused.

3. Review findings

The skill reads each guide’s review instructions and systematically examines the scripts. Findings are presented grouped by guide and severity:

Shell Scripting (3 findings)
  High: .aitask-scripts/aitask_sync.sh:42 — Unquoted variable in rm command risks glob expansion
  Medium: .aitask-scripts/aitask_sync.sh:15 — Missing set -euo pipefail
  Low: .aitask-scripts/aitask_helper.sh:88 — Hardcoded /tmp path instead of mktemp

Error Handling (1 finding)
  Medium: .aitask-scripts/aitask_sync.sh:67 — Broad trap on EXIT masks specific error handling

4. Select and create a task

You select the high and both medium findings (skip the low cosmetic one). Choose “Single task” — a task is created with the three findings in its description, priority set to high (from the highest severity finding).

Select “Continue to implementation” to hand off to the standard /aitask-pick workflow, which handles planning, implementation, review, and archival.

Managing Review Guides

Review guides live in aireviewguides/ organized by environment (general/, python/, shell/, etc.). Each guide is a markdown file with metadata and actionable review instructions. Guides can be excluded from auto-detection by adding patterns to aireviewguides/.reviewguidesignore (uses gitignore syntax). Three companion skills manage the guide library:

Typical guide management workflow:

1. Import — Use /aitask-reviewguide-import to bring in external coding standards, style guides, or best practices from URLs, local files, or GitHub repositories. The skill transforms narrative content into actionable review checklists
2. Classify — Use /aitask-reviewguide-classify to assign metadata (type, labels, environment) to new or unclassified guides. This metadata powers the auto-detection when running reviews
3. Merge — Use /aitask-reviewguide-merge to consolidate overlapping guides. When classify finds similar guides, merge resolves the overlap by combining content or deduplicating shared checks

When to Review

• After implementing a feature — Review the implementation for quality, security, and adherence to project conventions
• After a batch of changes — Select recent commits and review them collectively
• Periodic codebase audits — Review specific directories or modules against relevant guides
• Onboarding new standards — Import a new coding standard as a review guide, then review existing code against it to create tasks for bringing the codebase into compliance

Tips

• Keep reviews focused — Select one or two guides per session rather than running all guides at once. Focused reviews produce actionable findings; broad reviews produce noise
• Review after implementing — Use “Recent changes” to review your own work before moving on. This catches issues while context is fresh
• Build project-specific guides — Use /aitask-reviewguide-import to turn your team’s coding standards or style guides into review guides. Project-specific conventions are where automated review adds the most value
• Pre-configure frequent combinations — Set the review_default_modes profile key (comma-separated guide names) to auto-select your go-to guides, skipping the selection prompt