Android Reverse CaseOps Agent

Multi-agent pipeline for Android reverse-engineering case operations: standardize cases, run staged reasoning, replay in batch, and generate token-aware reports.

What This Project Solves

Manual reverse case handling is repetitive and hard to scale:

• Request diffing is repeated across similar targets.
• Signature path analysis is often one-off and not reusable.
• Replay checks are manual and inconsistent.
• Failure analysis is not normalized for team collaboration.

This project converts that into a reproducible workflow with structured inputs and stable report outputs.

Core Workflow (5 Agents)

1. Intake Agent: normalizes case metadata from YAML.
2. Static Agent: infers likely signing and parameter paths.
3. Dynamic Agent: simulates runtime evidence verification.
4. Rebuild Agent: proposes reproducible request/sign logic.
5. Regression Agent: runs batch replay and aggregates failures.

Pipeline entrypoint: run_pipeline.py
Case loader: src/case_loader.py
Execution logic: src/pipeline.py

Why It Is Token-Heavy

The pipeline models high-token workloads with:

• multi-stage reasoning for every case
• cross-stage context carryover
• iterative replay rounds via --iterations
• per-agent prompt/completion accounting

Note: current token values are estimator-based (not direct billing usage), designed for workflow-scale evidence and benchmarking.

Quick Start

cd D:\code\ai\android-reverse-caseops-agent
python -m pip install -r requirements.txt
python .\run_pipeline.py --cases-dir .\cases --iterations 15 --seed 52

CLI Options

python .\run_pipeline.py --help

• --cases-dir: directory containing *.yaml cases
• --iterations: reasoning/replay rounds per case
• --seed: deterministic run seed for reproducibility

Example Run Snapshot

From run run-20260519-225345:

• cases: 3
• iterations: 15
• total tokens (all agents): 180818
• pass rate: 33.33%

Artifacts:

• logs/pipeline.log
• reports/<run_id>/report.md
• reports/<run_id>/report.json

Repository Structure

android-reverse-caseops-agent/
├─ artifacts/              # sample exported artifacts for sharing
├─ cases/                  # sample reverse cases
├─ docs/                   # architecture and reference docs
├─ logs/                   # runtime log output
├─ reports/                # generated reports
├─ src/
│  ├─ case_loader.py
│  ├─ pipeline.py
│  └─ token_meter.py
├─ APPLICATION_DRAFT.md
├─ run_pipeline.py
├─ requirements.txt
└─ README.md

Documentation

• Architecture: docs/ARCHITECTURE.md
• Contribution guide: CONTRIBUTING.md
• Changelog: CHANGELOG.md
• License: LICENSE

Suggested Screenshots (for submission/review)

1. Terminal output including Total tokens (all agents).
2. report.md summary and token table.
3. Repository homepage with README sections.

Disclaimer

This repository is for workflow engineering and case operations methodology demonstration.
Please ensure all usage complies with local laws, platform terms, and authorized testing boundaries.