Development Setup

Prerequisites

• bash 4.0+
• python3
• git
• ShellCheck (for linting)

Clone the repo

git clone https://github.com/ugiordan/adversarial-reviewing.git
cd adversarial-reviewing

Repository structure

adversarial-reviewing/
  .claude-plugin/           # Marketplace metadata
  .cursor/rules/            # Cursor IDE rules
  .github/workflows/        # CI (test.yml)
  adversarial-review/       # Plugin package
    .claude-plugin/
      plugin.json           # Plugin manifest
    commands/
      adversarial-reviewing.md # Command definition
    skills/adversarial-reviewing/
      SKILL.md              # Main orchestration procedure
      profiles/             # Code and strategy profiles
      phases/               # Phase procedures
      protocols/            # Operational protocols
      scripts/              # Validation and utility scripts
      tests/                # Test suite
      references/           # Knowledge base modules
      templates/            # Output templates (legacy)
  docs/                     # Documentation and specs
  AGENTS.md                 # Universal agent prompts
  README.md
  Makefile

Running tests

cd adversarial-review/skills/adversarial-reviewing
bash tests/run-all-tests.sh

The test suite covers:

• Output validation (structure, injection detection)
• Triage validation
• Convergence detection
• Budget tracking
• Deduplication
• Reference module discovery
• Cache management
• Context fetching
• Diff and impact graph
• Comment parsing
• Guardrail enforcement
• Single-agent pipeline integration

Linting

ShellCheck runs in CI. To run locally:

shellcheck scripts/*.sh tests/test-*.sh

Suppressed checks: SC2329, SC2317 (unreachable code in test functions), SC1091 (source paths).

CI pipeline

The .github/workflows/test.yml workflow:

1. Test job: Runs tests/run-all-tests.sh, asserts 0 failures
2. ShellCheck job: Lints all bash scripts

Local testing with Claude Code

To test changes locally before pushing:

1. Build and install the plugin from your working copy:

   rsync -a adversarial-review/ \
     $HOME/.claude/plugins/cache/ugiordan-adversarial-reviewing/adversarial-review/1.0.0/
   

2. Start a new Claude Code session

3. Run /adversarial-reviewing against test code

Writing tests

Test scripts follow the pattern:

#!/usr/bin/env bash
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
source "$SCRIPT_DIR/../scripts/some-script.sh"

test_some_behavior() {
  local result
  result=$(some_function "input")
  [[ "$result" == "expected" ]] || { echo "FAIL: test_some_behavior"; return 1; }
  echo "PASS: test_some_behavior"
}

test_some_behavior

Add new test files to tests/ and they'll be picked up by run-all-tests.sh automatically.