code-review

Category: Development & Coding | Uploader: Shubchynskyi | Downloads: 0 | Version: v1.0（Latest）

Independent runtime code review with strict pass/fail verdict and auditable evidence. Use for requests like "code review", "review this diff", "review PR", "review before merge", or when preflight requires code review. Do NOT use for architecture brainstorming without concrete code changes.

Changelog: Source: GitHub https://github.com/Shubchynskyi/Octopus-agent-orchestrator

Directory Structure

Current level: template/skills/code-review/

📁 agents/
- 📄 openai.yaml 322 B
📁 references/
- 📄 code-review-checklist.md 1.5 KB
📄 SKILL.md 4.2 KB

SKILL.md

---
name: code-review
description: Independent runtime code review with strict pass/fail verdict and auditable evidence. Use for requests like "code review", "review this diff", "review PR", "review before merge", or when preflight requires code review. Do NOT use for architecture brainstorming without concrete code changes.
allowed-tools:
  - Read
  - Grep
  - Glob
  - Bash(pwsh:*)
  - Write
metadata:
  author: Octopus-agent-orchestrator
  version: 1.3.0
  runtime_requirement: PowerShell 7+ (pwsh) for gate scripts
---

# Code Review

Use this skill to produce a release-blocking technical review.
Review for defects and risks first. Keep summary secondary.

## Required Inputs
- Task goal and expected behavior.
- Changed files list.
- Diff summary or patch.
- Optional review-context artifact from orchestration: `Octopus-agent-orchestrator/runtime/reviews/<task-id>-<review-type>-review-context.json`.
- Rule context package selected by orchestration and explicitly passed to reviewer:
  - token economy active + `depth=1`: only `00-core.md`, `80-task-workflow.md`, and rule ids/snippets directly triggered by changed scope.
  - token economy active + `depth=2`: `00-core.md`, `35-strict-coding-rules.md`, `50-structure-and-docs.md`, `70-security.md`, `80-task-workflow.md`.
  - token economy disabled or `depth=3`: full required rule set for code review scope.

## Token Economy Mode
- Config source: `Octopus-agent-orchestrator/live/config/token-economy.json`.
- Apply this section only when `enabled=true` and effective depth is in `enabled_depths`.
- While active, this section takes precedence over any static rule-file list in `Required Inputs`.
- If orchestration provides review-context artifact, treat its `rule_pack.selected_rule_files`, `rule_pack.omitted_rule_files`, and `token_economy.omitted_sections` as the source of truth for compact review scope and omission evidence.
- Depth-aware required-rules behavior:
  - `depth=1`: evaluate only required rules directly triggered by changed scope; do not request full static rule bundle; list deferred required rules in `not_applicable_rule_ids` with reason `deferred_by_depth`.
  - `depth=2`: evaluate all required code-review rules for changed scope.
  - other depths: follow full review behavior without token-economy reductions.
- Compact mode contract:
  - when `compact_reviewer_output=true`, keep the same mandatory output sections and exact verdict token.
  - keep findings concise (`risk -> evidence -> required action`) and move detail overflow to residual risks.
  - when including failing command/test snippets, cap pasted tail output to `fail_tail_lines`.

## Review Workflow
1. Build scope from changed files and diff.
2. Load checklist from `references/code-review-checklist.md`.
3. Validate correctness, regressions, edge cases, and security impact.
4. Validate test coverage adequacy for changed behavior.
5. Validate documentation impact handling and required doc updates.
6. Validate rule compliance using rule ids and evidence.
7. Use artifact structure from `Octopus-agent-orchestrator/live/docs/reviews/TEMPLATE.md`.
8. Produce final verdict.

## Mandatory Output Format
1. Findings by severity with file references.
2. Rule checklist rows with `rule_id`, `status` (`PASS` or `FAIL`), `evidence`.
3. Rule coverage declaration with `applicable_rule_ids`, `not_applicable_rule_ids`, and reason for each skipped rule id.
4. Residual risks and testing gaps.
5. Explicit verdict: `REVIEW PASSED` or `REVIEW FAILED`.

## Hard Fail Conditions
Return `REVIEW FAILED` when any item is true:
- Unresolved critical or high-severity finding exists.
- Required tests are missing for runtime behavior changes.
- Rule checklist has `FAIL` without approved exception artifact.
- Rule checklist or coverage declaration is incomplete for applicable non-automated rules.
- Evidence is missing or non-auditable.

## Evidence Rules
- Use file references with line numbers when possible.
- If referencing command checks, include exact command and key output snippet.
- If exception is used, include the exception artifact location and rule id.

## Escalation
- Escalation triggers are defined only in `Octopus-agent-orchestrator/live/skills/orchestration/references/review-trigger-matrix.md`.
- Do not duplicate trigger rules in this skill.

Comments 0

Latest | Hot

Please login before commenting.

No comments yet. Be the first one!

code-review

Directory Structure

SKILL.md

Report

Notice