Battleprint - Agentic Execution of Red and Blue Team Security Analysis

You are executing the Battleprint skill. This performs a red and blue team adversarial analysis sequence.

Architecture Overview

This skill is a prompt-first architecture where YOU (Claude Code) handle all reasoning and analysis, while Python utilities provide fast data operations (file scanning, use of ai-fingerprint skill and its results).

Your Task

Execute a 4-stage analysis process that produces:

1. Red Stage 1: Complete red team analysis of the target code base
2. Blue Stage 1: Complete blue team analysis of the target code base
3. Red Stage 2: Utilizes Blue Team's analysis to refine red team's attack plans
4. Blue Stage 2: Utilizes Red Team's analysis to refine blue team's defense plans

Arguments

The user provides a target directory path.

Expected format: <target_dir>

Examples:

• ./my-app — analyze my-app directory
• /bob/fish — analyze absolute path

Required Skills

• ai-fingerprint: Generates file_scan.json and system_analysis.json for the same target directory
  • Run it first so this skill can reuse its outputs.
  • Output location: <parent>/ai_fingerprint_results/<target_name>/

Execution Steps

EXECUTION GUIDE - APPLIES TO ALL PHASES AND STEPS

When executing bash or unix (e.g. ls, wc, cd, etc.) commands, never ever string them together into a single command.

• Never use &&
• Never use ||
• Never use compound statements, ever
• Always execute a bash command one at a time.
• Always execute a unix command one-at-a-time, never compound them.

When generating JSON, never ever output it directly. Always write python code that generates the json

• Never generate JSON directly, it's too difficult.
• Always write the python code that generates the JSON.
• Always use python code generator to generate JSON.
• Always use python validation code, to read and confirm JSON was generated properly.
• Take your time. Generate Python to output JSON. Generate Python to Validate JSON.

Phase 0: Setup and Check for AI-FingerPrint

Parse and set paths:

# Get absolute path of target
TARGET_DIR=$(cd "$target_dir" && pwd)
TARGET_BASE=$(dirname "$TARGET_DIR")
TARGET_NAME=$(basename "$TARGET_DIR")

# Fingerprint results location (alongside target)
FINGERPRINT_DIR="${TARGET_BASE}/ai_fingerprint_results/${TARGET_NAME}"

# Battleprint output location (alongside target)
OUTPUT_DIR="${TARGET_BASE}/battleprint_results/${TARGET_NAME}"
mkdir -p "$OUTPUT_DIR"

All subsequent references to paths use these variables.

Verify fingerprint exists:

1. Check if $FINGERPRINT_DIR/system_analysis.json exists
2. If it does NOT exist, invoke the ai-fingerprint skill with $TARGET_DIR as argument. Use a SEPARATE AGENT (via Agent tool) for this to prevent context bleed.
3. Read $FINGERPRINT_DIR/file_scan.json and $FINGERPRINT_DIR/system_analysis.json into the current context.

Phase 1: Stage 1 Red and Blue

Objective: Perform Red Team and Blue Team Analysis on the target directory.

Process:

1. Read the file scan results:

cat "$FINGERPRINT_DIR/file_scan.json"

2. Read the system analysis results:

cat "$FINGERPRINT_DIR/system_analysis.json"

3. Parallel Execute 2 Agents

   Use the Agent tool to spawn two agents in parallel. Each Agent invocation is automatically in fresh context.

   a. Red Team Agent - use prompts/red_team_stage1.md with the fingerprint data b. Blue Team Agent - use prompts/blue_team_stage1.md with the fingerprint data

   Each agent should write its output to:

   • Red: $OUTPUT_DIR/red_stage_1.json
   • Blue: $OUTPUT_DIR/blue_stage_1.json
4. Verify Stage 1 Complete

Use Python's built-in JSON parser to check validity:

python3 -c "import json; json.load(open('$OUTPUT_DIR/red_stage_1.json'))" && echo "✓ Valid JSON" || echo "✗ Invalid JSON"

and

python3 -c "import json; json.load(open('$OUTPUT_DIR/blue_stage_1.json'))" && echo "✓ Valid JSON" || echo "✗ Invalid JSON"

Phase 2: Stage 2 Red and Blue

Objective: Perform a second pass Red and Blue Team Analysis on the target directory.

Note: Each Agent tool invocation automatically runs in fresh context - no special action needed.

Red Team Agent (Stage 2)

1. Read the file scan results:

cat "$FINGERPRINT_DIR/file_scan.json"

2. Read the system analysis results:

cat "$FINGERPRINT_DIR/system_analysis.json"

3. Read blue team's plans:

cat "$OUTPUT_DIR/blue_stage_1.json"

4. Execute Red Team Stage 2

   Use Agent tool with prompts/red_team_stage2.md and the above context.

   Write output to: $OUTPUT_DIR/red_stage_2.json

5. Verify Stage 2 Complete

python3 -c "import json; json.load(open('$OUTPUT_DIR/red_stage_2.json'))" && echo "✓ Valid JSON" || echo "✗ Invalid JSON"

Blue Team Agent (Stage 2)

1. Read the file scan results:

cat "$FINGERPRINT_DIR/file_scan.json"

2. Read the system analysis results:

cat "$FINGERPRINT_DIR/system_analysis.json"

3. Read red team's plans:

cat "$OUTPUT_DIR/red_stage_1.json"

4. Execute Blue Team Stage 2

   Use Agent tool with prompts/blue_team_stage2.md and the above context.

   Write output to: $OUTPUT_DIR/blue_stage_2.json

5. Verify Stage 2 Complete

python3 -c "import json; json.load(open('$OUTPUT_DIR/blue_stage_2.json'))" && echo "✓ Valid JSON" || echo "✗ Invalid JSON"

Phase 3: Synthesis

Perform synthesis and provide final results:

1. Output a final_analysis.json with the top 8 exploits found in red_team 2. Did blue_team 2 have an answer for these?
2. Provide an analysis of the efficacy of this battleprint algorithm: a. Did the blue team improve from stage 1 to stage 2 (i.e. did it utilize the knowledge it gained by reading red team plans and improve its defenses against these?). If so, how? b. Did the red team improve from stage 1 to stage 2? If so, how?
3. Provide final recommendations (top 3) for the engineering team to address. When providing these top 3, be sure to utilize the system analysis to note which AI System and sub-system are most affected.
4. Provide a weakness indicator for each element in the strand. Use relative weakness (i.e. element with the highest relative weakness gets a 10, element with no relative weakness gets a 0).

All output should be in 2 forms:

1. JSON file: $OUTPUT_DIR/final_analysis.json (enables downstream processing)
2. Markdown file: $OUTPUT_DIR/final_analysis.md (allows for easy readability)

Output location: Results are saved alongside the target directory at: <parent_of_target>/battleprint_results/<target_name>/