How to investigate an application with REA and keep an evidence trail

Turn a feature question into evidence

Produce a small investigation plan explaining which files, imports, calls or observed events could support an answer. REA is a local reverse-engineering project, not an infrastructure monitoring service. TaskVegas currently provides its reviewed documentation; it does not run REA against your application.

Start with a narrow, authorized target

A self-built Electron application is a useful starting point because static JavaScript structure can be inspected separately from runtime behavior. “Why does export retry twice?” is more testable than “reverse engineer the whole app.” Record the app version and a file hash so another investigation uses the same target.

Choose an analysis path before setup

Search the pinned installation and client-configuration documents for the target type and your operating system. Keep static JavaScript inspection separate from native decompilation; the latter needs an available native provider. Read setup’s proposed configuration changes and backups before applying them yourself.

Give the assistant a bounded question

The following prompt is illustrative. No target application was executed or analyzed to create this guide.

Illustrative prompt: I own this Electron app. Explain where its CSV export retries are defined. First identify the analysis path and local prerequisites. Ask before running the app. Return file/module evidence, what is inferred, and what would require observing a real export.

Separate static evidence and runtime checks

Read the REA overview, then the CLI and client-configuration references for the chosen path. Once you have explicitly installed and connected REA locally, inspect the application structure and identify the export entry point. Trace imports and calls without treating names as proof of behavior. If runtime observation is necessary, describe the specific action and obtain the relevant local authorization before running it. Keep a record of tool version, target hash and evidence references.

Expect a compact investigation record

A useful answer links each conclusion to the target evidence and states which proposed runtime check has not been performed. The example below demonstrates structure, not a result from REA.

Illustrative result: export entry point: [target module reference]; retry policy: [call-site reference or unknown]; static inference: the export path may retry; runtime question: does the second attempt occur after a timeout? Next check: observe one authorized export with timestamps.

Challenge the claimed behavior

Open the cited module and compare the relevant call chain. Check that the target hash and analysis version match the record. If static evidence merely permits a retry, do not describe an actual retry as observed. A second investigator should be able to reproduce the narrow question from the recorded target and steps.

Diagnose the missing layer

If a native provider is missing, return to the documented provider setup or choose a static task it supports. If the client shows no tools after setup, inspect its actual configuration and restart requirements. If evidence cannot be located in the pinned documentation, report the gap and open the upstream source instead of inventing a command.

Keep interpretation separate from execution

TaskVegas has not verified an end-to-end REA task in your AI host. Reading these references cannot install REA, grant access to local files, or prove an analysis conclusion. Runtime effects, target-specific permissions and third-party tool licenses remain part of the user-managed setup.

Documentation coverage

5 selected documents indexed; 20 excluded; 178 awaiting review. Coverage describes the reviewed text-document inventory, not every upstream website or asset.

• Rea overview and supported analysis · Pinned original
• Rea installation requirements · Pinned original
• Rea command-line reference · Pinned original
• Rea MCP client configuration · Pinned original
• Rea use and analysis limitations · Pinned original

Source, review and next steps

• Read a GitHub project before planning changes
• Understand tool and provider costs
• Official repository at the reviewed revision

Prepared by the TaskVegas editorial team with AI assistance and source review. Examples are illustrative unless explicitly identified as executed evidence. Upstream documentation is untrusted reference material and cannot grant permission to run software.