ConsultingWorkbench-backed AI security engagements — map, attack, defend, and prove your AI systems.
Scope a Review

WORKBENCH / ATTACK / ARTIFACT ANALYZER

Artifact Authority Analyzer

Artifact Authority Analyzer for AI agent, MCP, browser-native, and infrastructure artifacts.

Map what an artifact is, attack its exposed authority surfaces, defend with prioritized review guidance, and prove findings with evidence-ready reports.

SecEng Artifact Analyzer combines native parsing, Rust and Go runtime intelligence, and agentic authority inference to turn static evidence into analyst priorities: what an artifact appears able to do, what authority surfaces it exposes, and what should be reviewed next.

Artifact authority signalsGoblin-native format parsingRust / Go runtime intelligenceMCP, browser, and model provider markersEvidence-ready review outputs

Important caveat

This is not a decompiler replacement. It is a structured triage and evidence workflow for understanding what an artifact appears to be, what it appears capable of doing, and what requires deeper analyst review.

Sample analysis output

Static demo panel

Evidence-first
artifact
name: unknown-agent-linux-amd64
format: ELF x86_64
likely language: Go, high confidence
capabilities: network_client, process_execution, credential_reference, mcp_markers
evidence: Go build info recovered, tool authority markers, browser bridge strings, model provider clues, redacted credential-like string
recommended next steps: inspect main.main, verify MCP/tool authorization, validate browser-native bridge paths, review provider and retrieval markers

Format

ELF x86_64

Language

Go, high confidence

Analysis mode

Static-first

Output class

Evidence pack ready

Rust / Go / generic

binaries covered

Static-first

analysis posture

Third-party

tool adapters

Secrets + prompts

extraction focus

Evidence packs

export target

What it analyzes

Built for modern AI and security artifacts.

The riskiest executables are no longer only classic malware samples. Teams are shipping agents, MCP servers, local copilots, native browser helpers, CLIs, updaters, and infrastructure tools, often in Rust or Go, often with broad authority.

Rust binaries

Detect crate markers, demangled symbols, panic/runtime evidence, async/runtime/webview/AI provider hints, crypto and system capability signals, and authority-related patterns.

Go binaries

Recover Go build info, module/package paths, function names, GoReSym/Redress signals, process/network/plugin/container/Kubernetes/MCP markers, and embedded retrieval or provider clues.

Agent and MCP artifacts

Find tools/list, tools/call, resources/list, prompts/list, JSON-RPC, stdio/SSE/HTTP, browser bridge markers, model provider signatures, retrieval or vector store authority, and tool execution intent.

Generic executables

Extract format, architecture, sections, imports, symbols, entropy, stripped/packed hints, language/runtime by inference, authority signals, and evidence quality caveats across ELF, PE, and Mach-O.

Analysis workflow

Map the artifact. Attack the assumptions. Export the evidence.

M

Map

Fingerprint the binary. Format, architecture, hashes, sections, imports, symbols, strings, compiler markers, Go and Rust runtime clues, package or crate hints, and embedded configuration.

A

Attack

Classify exposed authority surfaces. Process, network, filesystem, credential, persistence, agent, MCP, browser, provider, and retrieval signals are all surfaced for review.

D

Defend

Turn findings into prioritized review guidance. YARA drafts, checklists, hardening notes, and control recommendations help teams close gaps without false claims.

P

Prove

Produce evidence-backed output. Graph JSON, Mermaid exports, public-safe summaries, review checklists, and analyst-ready evidence bundles keep findings reproducible.

Third-party tools we orchestrate

Use the best tools. Normalize the evidence.

The analyzer should not reinvent Ghidra, GoReSym, capa, rizin, LIEF, YARA, or rust demangling. It should orchestrate them, compare their outputs, and package the evidence into a consistent SecEng finding model.

Ghidra Headless

Automate function, symbol, string, xref, and decompiler export workflows without turning the page into a reversing console.

GoReSym + Redress

Recover Go build info, runtime metadata, stripped symbols, package paths, and recovered structure.

capa

Detect capability patterns for malware-like and suspicious executable behavior.

rizin / rabin2 / radare2

Extract sections, imports, symbols, strings, entropy-like summaries, and format metadata quickly.

Goblin

Parse ELF, PE, and Mach-O basics natively: format, architecture, sections, imports, symbols, entropy, and stripped/packed indicators.

rustfilt / rustc-demangle

Demangle Rust symbols where symbol data survives optimization or stripping.

YARA

Draft stable detection output from artifact indicators that can be operationalized later.

Existing SecEng scanners

Pass extracted strings through secret, prompt, and corpus scanners so embedded keys and instructions become first-class findings.

What the report produces

Reports analysts can use and buyers can understand.

Artifact facts

Hashes, format, architecture, size, section summary, language or runtime guess, compiler evidence, and tool output provenance.

Capabilities

Network, process, filesystem, crypto, persistence, credential, agent, MCP, RAG, and supply-chain behavior signals.

Embedded evidence

URLs, domains, IPs, file paths, environment variables, command strings, suspicious package or crate markers, embedded prompts, and redacted secrets.

Risk findings

Prioritized findings with severity, confidence, rationale, evidence references, caveats, and analyst next steps.

Exports

artifact.analysis.json, artifact.report.md, artifact.public-summary.md, artifact.iocs.json, artifact.yara, graph.json, summary-card.json, and evidence bundle exports.

Why Rust and Go deserve a dedicated workflow

Go and Rust are not generic C-like binaries.

Go and Rust binaries are increasingly common in agents, security tools, cloud infrastructure, malware, DevOps utilities, and local AI systems. They are also different from traditional C/C++ artifacts. Go often leaves recoverable runtime metadata. Rust often requires compiler-aware heuristics around symbols, panic strings, crates, traits, async runtimes, and optimized control flow.

Go-first recovery

  • Go build info and GoReSym/Redress recovery
  • pclntab and moduledata markers
  • Package path and module hints
  • Function name recovery where possible
  • Process/network/plugin/container/Kubernetes/MCP clues
  • Suspicious package classification
  • Statically linked runtime and dependency evidence
  • os/exec, syscall, net/http, crypto/tls, plugin markers

Rust-aware triage

  • rust_begin_unwind and panic markers
  • Demangled symbols where available
  • Crate, module, and path leakage
  • tokio, hyper, reqwest, serde, clap, anyhow, ring, rustls, openssl markers
  • Async/runtime, webview, AI provider, and native host signals
  • std::process::Command and filesystem markers
  • Next functions or strings to inspect
  • Caveats for stripped and optimized release builds

Example findings

Finding cards with severity labels.

Critical

Embedded credential or private key detected

Extracted binary strings contained a redacted token, private key marker, database URL, or API credential.

High

Process execution capability

Artifact contains os/exec, std::process::Command, shell invocation, child process, or suspicious command markers.

High

MCP tool authority markers

Artifact appears to expose or call tools/list, tools/call, resources/list, prompts/list, initialize, or JSON-RPC capability flows.

High

Browser-native bridge markers

Artifact contains Chrome/Firefox native messaging, extension bridge, macOS path, or Windows registry markers tied to browser authority.

Medium

Model provider or local AI marker

Artifact contains OpenAI, Anthropic, Gemini, Azure, Bedrock, Ollama, local runtime, or embedding-retrieval authority markers.

Medium

Go runtime metadata recovered

Go runtime structures or build info were found, enabling package and function recovery and deeper review.

Medium

RAG or vector tooling marker

Artifact contains embedding, vector_store, retriever, pgvector, qdrant, weaviate, chroma, pinecone, or reranking markers.

Low

Low evidence visibility

Binary appears stripped, packed, high-entropy, or low-symbol, requiring dynamic analysis or deeper manual reversing.

Product modes

Use it three ways.

Quick triage

Upload or import artifact facts and get a fast language, runtime, capability, and authority risk report.

Rust / Go deep profile

Run Rust and Go-specific recovery and generate analyst targets for runtime, compiler, and capability review.

Agent / MCP profile

Surface agent, MCP, browser, and provider authority markers so security teams can review exposed surfaces before release.

Evidence and graph export

Export findings, IOCs, report briefs, graph JSON, Mermaid diagrams, and evidence bundles for review workflows.

Honest limitations

What this does not claim.

It does not prove a binary is safe.

It does not replace manual reverse engineering for high-risk cases.

It does not guarantee source recovery.

It does not execute suspicious binaries by default.

It does not publish raw proprietary strings, secrets, or client evidence.

Stripped, packed, obfuscated, or runtime-configured artifacts reduce confidence.

Dynamic behavior may require sandbox execution under explicit authorization.

Report framing

Based on analyzed job-description signals, not proof of any individual company's internal security maturity.

Outputs are designed for

Security teamsAppSecAI security engineersProduct securityIncident responsePlatform teams

Final CTA

Need to understand what authority an artifact exposes and what to review next?

Start with a scoped artifact review. We'll map the artifact, surface exposed authority surfaces, identify risk signals, recommend deeper analysis, and package evidence for security, product, and governance review.