{
  "id": "reverse-engineering-101",
  "title": "Reverse Engineering 101",
  "subtitle": "Reading a compiled program you built, from file bytes to evidence-backed pseudocode, with rea and Ghidra",
  "summary": "How to read a compiled program without its source and prove what you found. Every listing comes from a recorded run of rea 6.7.0 and Ghidra 12.1.4 on programs written for this manual.",
  "audience": "You are an engineer who writes C, Python, or JavaScript and uses a coding agent. You have never read machine code with a decompiler. You want to understand what a compiled program does and to check what an agent tells you about it.",
  "edition": "2026.10",
  "status": "ready",
  "pin": {
    "subject": "rea (rea-agents) with the Ghidra provider",
    "version": "rea-agents 6.7.0, Ghidra 12.1.4",
    "source": "https://github.com/morluto/rea",
    "commit": "27dc42d10249f0410904a75c0997842eb03b931c",
    "date": "2026-10-18",
    "verified": "2026-10-18"
  },
  "outcomes": [
    "Read the file header, sections, symbols, and relocations of a compiled program, and map an address to a file offset.",
    "Run rea with the Ghidra provider on a program you built, and read its Evidence record: confidence, authority, limitations, and unknowns.",
    "Find a function, read its assembly and pseudocode, follow its references, and say where the decompiler output is incomplete.",
    "Prove a rebuilt function equal to the original by running both over an input grid and by comparing bytes after masking relocations.",
    "Drive rea from an agent over MCP, and keep an evidence ledger that separates what was observed from what was inferred."
  ],
  "palette": [
    {
      "hue": "blue",
      "meaning": "rea: the CLI, the MCP server, and its domain layer"
    },
    {
      "hue": "violet",
      "meaning": "the Ghidra provider: analyzeHeadless, the bridge, and the decompiler"
    },
    {
      "hue": "green",
      "meaning": "the artifact: the fixture files, their bytes, sections, and symbols"
    },
    {
      "hue": "plum",
      "meaning": "the agent and its model calls"
    },
    {
      "hue": "olive",
      "meaning": "scripted and managed code: the JavaScript app and the WASM module"
    },
    {
      "hue": "indigo",
      "meaning": "MCP messages and sessions"
    },
    {
      "hue": "amber",
      "meaning": "lifecycle: sessions, analysis phases, and claim states"
    },
    {
      "hue": "teal",
      "meaning": "durable records: Evidence bundles and snapshots"
    },
    {
      "hue": "rose",
      "meaning": "failure: error envelopes, refused claims, and wrong output"
    },
    {
      "hue": "grey",
      "meaning": "outside rea: the host, the compiler, and the reader"
    }
  ],
  "sources": [
    {
      "name": "rea source at the release tag",
      "path": "rea-agents-6.7.0 (27dc42d)",
      "usedFor": "every claim about how rea works; files listed in each Sources line"
    },
    {
      "name": "rea documentation at the release tag",
      "path": "docs/ and DISCLAIMER.md at 27dc42d",
      "usedFor": "documented behaviour, limits, and the responsible use rules"
    },
    {
      "name": "Captured runs",
      "path": "capture/out/ (rea 6.7.0, Ghidra 12.1.4, macOS 26.2 arm64)",
      "usedFor": "every listing and every number in a figure"
    },
    {
      "name": "Format and architecture specifications",
      "path": "ELF gABI, Mach-O headers, Intel SDM, Arm ARM",
      "usedFor": "ELF, Mach-O, x86-64 and AArch64 encoding, and calling conventions"
    },
    {
      "name": "Papers on program analysis",
      "path": "papers named in each Sources line",
      "usedFor": "disassembly, control-flow structuring, SSA, type recovery, and decompiler evaluation"
    },
    {
      "name": "Statutes and court decisions",
      "path": "17 U.S.C. 1201, Directive 2009/24/EC, CJEU and US decisions",
      "usedFor": "the law of analysis and interoperability"
    },
    {
      "name": "Tool release notes",
      "path": "vendor release notes named in each Sources line",
      "usedFor": "Ghidra and other tool versions and features"
    }
  ],
  "quoteSources": {
    "readme": null,
    "cli": null,
    "mcp": null,
    "ghidra": null,
    "native": null,
    "install": null,
    "readiness": null,
    "ledgers": null,
    "jsgraph": null,
    "design": null,
    "law": null,
    "spec": null,
    "paper": null
  },
  "capture": {
    "run": [
      "python3",
      "capture/run.py"
    ],
    "check": [
      "python3",
      "capture/run.py",
      "--check"
    ],
    "timeoutSeconds": 900
  },
  "front": "front.md",
  "plate": {
    "figure": "plate-rea-investigation",
    "title": "One function, four views",
    "caption": "The C you wrote, the six instructions clang made, the pseudocode Ghidra returned, and the Evidence record rea wrote for them, from capture/out/06-function-percent-change.json."
  },
  "parts": [
    {
      "number": 1,
      "title": "rea on One Page",
      "thesis": "rea asks an analysis engine about a program and returns each answer as an Evidence record that says how it knows.",
      "summary": "What rea standardizes, what it leaves to Ghidra, and one complete investigation of a function.",
      "accent": "blue",
      "sections": [
        {
          "id": "1.1",
          "file": "sections/1-1-rea-providers-and-evidence.md"
        },
        {
          "id": "1.2",
          "file": "sections/1-2-rea-function-end-to-end.md"
        }
      ]
    },
    {
      "number": 2,
      "title": "Authority and Setup",
      "thesis": "You analyze a program only when you have the right to, and only after rea doctor reports the provider healthy.",
      "summary": "The law and rea's responsible use rules, the Ghidra provider install, and the capabilities each command declares.",
      "accent": "grey",
      "sections": [
        {
          "id": "2.1",
          "file": "sections/2-1-the-readme-use-rule-and-the-law.md"
        },
        {
          "id": "2.2",
          "file": "sections/2-2-rea-doctor-and-the-ghidra-provider.md"
        },
        {
          "id": "2.3",
          "file": "sections/2-3-rea-capabilities.md"
        }
      ]
    },
    {
      "number": 3,
      "title": "The Artifact",
      "thesis": "Before any engine runs, rea reads the file itself: its format, its load image, and where each address lives on disk.",
      "summary": "inspect-artifact, the binary layout and Mach-O load image, address-to-file-offset, and the analyze overview.",
      "accent": "green",
      "sections": [
        {
          "id": "3.1",
          "file": "sections/3-1-inspect-artifact.md"
        },
        {
          "id": "3.2",
          "file": "sections/3-2-inspect-binary-layout-and-the-load-image.md"
        },
        {
          "id": "3.3",
          "file": "sections/3-3-address-to-file-offset.md"
        },
        {
          "id": "3.4",
          "file": "sections/3-4-analyze-overview.md"
        }
      ]
    },
    {
      "number": 4,
      "title": "The Ghidra Provider",
      "thesis": "Each question rea asks Ghidra becomes one bridge call into a fresh headless project, and the answer carries the limits of that engine.",
      "summary": "open_binary and the bridge, procedures, assembly, references and the jump table, strings, and pseudocode.",
      "accent": "violet",
      "sections": [
        {
          "id": "4.1",
          "file": "sections/4-1-open-binary-and-the-headless-bridge.md"
        },
        {
          "id": "4.2",
          "file": "sections/4-2-list-procedures-and-function-bodies.md"
        },
        {
          "id": "4.3",
          "file": "sections/4-3-procedure-assembly.md"
        },
        {
          "id": "4.4",
          "file": "sections/4-4-xrefs-and-the-jump-table.md"
        },
        {
          "id": "4.5",
          "file": "sections/4-5-list-strings.md"
        },
        {
          "id": "4.6",
          "file": "sections/4-6-analyze-function-and-pseudocode.md"
        }
      ]
    },
    {
      "number": 5,
      "title": "Evidence",
      "thesis": "An Evidence record names its subject, its engine, its confidence, and its limits, and its id changes when any of them changes.",
      "summary": "The Evidence record, residual unknowns, snapshots, bundles, and comparing two builds.",
      "accent": "teal",
      "sections": [
        {
          "id": "5.1",
          "file": "sections/5-1-the-evidence-record.md"
        },
        {
          "id": "5.2",
          "file": "sections/5-2-residual-unknowns-and-coverage.md"
        },
        {
          "id": "5.3",
          "file": "sections/5-3-snapshots-and-the-answer-cache.md"
        },
        {
          "id": "5.4",
          "file": "sections/5-4-evidence-bundles-export-and-import.md"
        },
        {
          "id": "5.5",
          "file": "sections/5-5-compare-and-compare-functions.md"
        }
      ]
    },
    {
      "number": 6,
      "title": "Scripted and Managed Code",
      "thesis": "For JavaScript, rea reads the files without running them and returns a graph of modules and their relationships.",
      "summary": "analyze-javascript-application and the application graph, and inspect-wasm-artifact with the WABT provider.",
      "accent": "olive",
      "sections": [
        {
          "id": "6.1",
          "file": "sections/6-1-analyze-javascript-application.md"
        },
        {
          "id": "6.2",
          "file": "sections/6-2-inspect-wasm-artifact.md"
        }
      ]
    },
    {
      "number": 7,
      "title": "Rebuild and Prove",
      "thesis": "rea records whether declared claims passed, so the proof that a rebuilt function is equal must come from your own tests.",
      "summary": "verify_reconstruction and its limits, and a harness that proves a rebuild equal by execution and by bytes.",
      "accent": "rose",
      "sections": [
        {
          "id": "7.1",
          "file": "sections/7-1-verify-reconstruction.md"
        },
        {
          "id": "7.2",
          "file": "sections/7-2-a-proof-harness-outside-rea.md"
        }
      ]
    },
    {
      "number": 8,
      "title": "Agents over MCP",
      "thesis": "An agent reaches rea through one MCP server with 207 tools, and its findings are only as good as the evidence it keeps.",
      "summary": "The MCP server, its tools and deadlines, and one agent investigation from question to evidence.",
      "accent": "indigo",
      "sections": [
        {
          "id": "8.1",
          "file": "sections/8-1-the-rea-mcp-server.md"
        },
        {
          "id": "8.2",
          "file": "sections/8-2-an-agent-investigation.md"
        }
      ]
    }
  ],
  "reference": {
    "title": "Reference",
    "accent": "grey",
    "thesis": "Lookup tables for every rea name in the manual.",
    "summary": "Commands and MCP tools, Evidence fields, errors seen in the captures, a glossary, the sources, and the figure index.",
    "sections": [
      {
        "id": "R.1",
        "file": "sections/r-1-commands-and-tools.md"
      },
      {
        "id": "R.2",
        "file": "sections/r-2-evidence-fields.md"
      },
      {
        "id": "R.3",
        "file": "sections/r-3-errors.md"
      },
      {
        "id": "R.4",
        "file": "sections/r-4-glossary.md"
      },
      {
        "id": "R.5",
        "file": "sections/r-5-sources.md"
      },
      {
        "id": "R.6",
        "file": "sections/r-6-figure-index.md"
      }
    ]
  },
  "related": [
    "phases/13-tools-and-protocols/32-program-anatomy-elf-from-scratch",
    "phases/13-tools-and-protocols/33-disassembly-and-control-flow",
    "phases/13-tools-and-protocols/34-lifting-to-pseudocode",
    "phases/13-tools-and-protocols/35-evidence-confidence-and-unknowns",
    "phases/13-tools-and-protocols/36-rebuild-and-prove-equal"
  ]
}