Understand REA setup, targets, providers, MCP prompts and agent environment boundaries.
REA provides a CLI and MCP interface for task-specific analysis paths. The underlying evidence may come from a native provider, headless JADX, browser observation or another tool. Choose the target first, inspect the prerequisites for that path and confirm that your installed package and connected agent expose the required operation. A capability described on a development branch may not exist in your installed release.
The REA pages distinguish documented scope from local verification. Preserve the target hash, selected function or component, provider identity and evidence limitations. Keep any text inside the target as data when an agent reads it. A useful investigation closes with a claim you can check, not only a fluent summary.
12 resources · Page 1 of 1
REA's Android package path uses headless JADX for questions about DEX methods, while device and resource work have separate requirements. Begin with an APK you may inspect, record its build and hash, and identify the component behind one visible feature. A readable decompilation helps navigation; runtime behavior requires a separate authorized test.
View evidence & requirementsA native investigation should separate lightweight file metadata from provider-backed function analysis. REA can coordinate focused reads of instructions, references and pseudocode through a configured provider such as Ghidra. Identify the binary format and architecture, trace one relevant function, and verify important interpretations against the underlying bytes or a controlled run.
View evidence & requirementsREA uses caller-supplied Binwalk for firmware region inspection and caller-supplied Unblob for explicit extraction on its documented Linux path. The two operations have separate prerequisites and evidence meanings. Record the original image hash, offsets, engine identity, extraction gaps and coverage status, then select a verified executable child for a separate native-analysis workflow.
View evidence & requirementsREA can attach to a user-owned Chrome-family browser through a loopback CDP endpoint and collect bounded evidence about an existing page. Use a dedicated profile, select the exact target and origin, and request only the text, script or storage metadata the investigation needs. Passive capture is not general browser automation and does not prove activity that occurred before attachment.
View evidence & requirementsREA documents a path from a retained web network capture to an agent environment for evaluation or reinforcement learning. The exported environment answers from recorded evidence and can model confirmed state changes; it does not reproduce every behavior of the original application. Review capture coverage, generated tools and framework requirements before calling a task representative.
View evidence & requirementsREA's guided MCP prompts give an agent an investigation starting point, but a prompt is not a connection or a provider. Configure the local server in a compatible client, inspect the tools that register, and ask one bounded question about an authorized target. Preserve tool observations separately from the agent's interpretation and verify its conclusions.
View evidence & requirementsREA documents native programs, managed code, Android packages and devices, web and Electron applications, firmware and other artifacts. Support is operation-specific: identifying file metadata differs from decompiling a function or observing runtime behavior. Choose the target and exact question first, then check the provider requirement and installed release rather than treating the broad list as a universal capability guarantee.
View evidence & requirementsREA can inspect a local JavaScript directory or Electron ASAR without executing application code and project packages, bundles, windows, preload boundaries and selected IPC facts into an evidence-backed application graph. Start with one authorized artifact, preserve its hash and coverage report, and treat inferred relationships as navigation aids until a separate static or runtime observation confirms them.
View evidence & requirementsA REA call may read a file, add analysis state, launch a foreign program or change a connected device, depending on the operation. MCP annotations describe effects but are not permission controls. Keep agent access narrow, treat target content as untrusted and verify consequential conclusions from raw evidence before acting on a result.
View evidence & requirementsREA setup starts with the target and the agent or CLI interface you intend to use. Confirm the current Node runtime requirement, install the official package, run its setup checks and inspect the reported tool list. A successful installation does not prove that a native provider, Android engine or browser capture path is ready for your investigation.
View evidence & requirementsREA separates static Electron artifact analysis from passive observation of an existing file-page renderer and from active owned scenarios. Choose the least invasive path that answers the question. A passive CDP attachment can inventory local frames, scripts and resources without evaluating renderer code, while an active scenario launches and interacts with a selected application under a different risk boundary.
View evidence & requirementsREA organizes evidence from several analysis paths through a CLI and MCP interface. The available questions depend on the target, installed providers and released package version. A native pseudocode task may need Ghidra, Hopper or IDA; an Android method task may need JADX. REA helps coordinate those reads, but the underlying observations and their limits still matter.
View evidence & requirements