Static binary analysis with angr: decompilation, CFG recovery, symbolic execution, vulnerability detection

GitHub stars
18

Install

/plugin install angr-analyzer@my-claude-skills

Run it inside Claude Code after adding the marketplace with /plugin marketplace add sandbornm/my-claude-skills.

Contents

Discovered from source

Skills 1

angr Binary Analysis

Perform automated binary analysis using angr's Python framework. Load binaries, recover control flow, decompile to C, extract symbols, and run symbolic execution to find bugs and solve for inputs.

When to Use

  • Decompiling a binary to C pseudocode (works on stripped binaries)
  • Recovering control flow graphs from executables
  • Finding vulnerabilities via symbolic execution and taint analysis
  • Extracting functions, strings, and symbols from binaries
  • Solving for inputs that reach specific code paths (CTF, crash analysis)
  • Cross-referencing functions and data in compiled code
  • Analyzing firmware or embedded binaries

When NOT to Use

  • Source code is available — read it directly
  • Interactive debugging needed — use GDB, LLDB
  • Dynamic runtime analysis needed — use Frida or a debugger
  • .NET assembly — use dnSpy or ILSpy
  • Java bytecode — use jadx or cfr
  • Quick string/symbol dump only — strings or nm may suffice

Quick Reference

TaskCommand
Full analysis{baseDir}/scripts/angr-analyze.sh -s analyze_all.py -o ./output binary
Decompile to C{baseDir}/scripts/angr-analyze.sh -s decompile.py -o ./output binary
List functions{baseDir}/scripts/angr-analyze.sh -s export_functions.py -o ./output binary
Extract strings{baseDir}/scripts/angr-analyze.sh -s export_strings.py -o ./output binary
Control flow graph{baseDir}/scripts/angr-analyze.sh -s export_cfg.py -o ./output binary
Export symbols{baseDir}/scripts/angr-analyze.sh -s export_symbols.py -o ./output binary
Cross-references{baseDir}/scripts/angr-analyze.sh -s export_xrefs.py -o ./output binary
Find vulnerabilities{baseDir}/scripts/angr-analyze.sh -s find_vulns.py -o ./output binary
Symbolic execution{baseDir}/scripts/angr-analyze.sh -s symbolic_explore.py -a "--find 0x401234 --avoid 0x401300" -o ./output binary

Prerequisites

  • Python 3.8+ with pip
  • angr: pip install angr

Main Wrapper Script

{baseDir}/scripts/angr-analyze.sh [options] <binary>

Options:

  • -o, --output <dir> — Output directory for results (default: current dir)
  • -s, --script <name> — Analysis script to run (can be repeated)
  • -a, --script-args <args> — Arguments for the last specified script
  • --base-addr <addr> — Base address for position-independent binaries
  • --auto-load-libs — Load shared library dependencies (default: off)
  • --timeout <seconds> — Analysis timeout
  • -v, --verbose — Verbose output
  • -h, --help — Show help

Built-in Analysis Scripts

analyze_all.py

Runs summary, decompilation, function listing, string extraction, and interesting-pattern detection. Best for initial triage.

Output files:

  • {name}_summary.txt — Architecture, entry point, sections, function counts
  • {name}_decompiled.c — All functions decompiled to C
  • {name}_functions.json — Function list with metadata
  • {name}_strings.json — Extracted strings with addresses
  • {name}_interesting.txt — Security-relevant function patterns

decompile.py

Decompile all recovered functions to C pseudocode using angr's decompiler.

Output: {name}_decompiled.c

export_functions.py

Export recovered functions as JSON with addresses, sizes, calling convention, and call targets.

Output: {name}_functions.json

export_strings.py

Extract ASCII and Unicode strings from binary sections.

Output: {name}_strings.json

export_cfg.py

Recover and export the control flow graph — angr's core strength. Includes nodes (functions/blocks), edges (calls/jumps), entry points, and call frequency analysis.

Output: {name}_cfg.json

export_symbols.py

Export all symbols, imports, exports, sections, and segments.

Output: {name}_symbols.json

export_xrefs.py

Build cross-reference maps: callers/callees for each function, plus data references.

Output: {name}_xrefs.json

find_vulns.py

Pattern-based vulnerability detection: dangerous function calls (strcpy, sprintf, gets, system), format string issues, and sensitive string references.

Output: {name}_vulns.json

symbolic_explore.py

Symbolic execution — angr's unique capability. Find inputs that reach a target address while avoiding others.

Arguments: --find <addr> --avoid <addr> [--avoid <addr>...] --timeout <secs>

Output: {name}_symbolic.json with found states, stdin, stdout, and constraint details.

Common Workflows

Triage an Unknown Binary

mkdir -p ./analysis
{baseDir}/scripts/angr-analyze.sh -s analyze_all.py -o ./analysis unknown_bin
cat ./analysis/unknown_bin_summary.txt
cat ./analysis/unknown_bin_interesting.txt

Find a Path to a Target Function (CTF)

# First find the target address
{baseDir}/scripts/angr-analyze.sh -s export_functions.py -o ./out binary
cat ./out/binary_functions.json | jq '.[] | select(.name=="win")'

# Then solve for input
{baseDir}/scripts/angr-analyze.sh \
    -s symbolic_explore.py \
    -a "--find 0x401234 --avoid 0x401300" \
    -o ./out binary
cat ./out/binary_symbolic.json

Analyze Firmware

{baseDir}/scripts/angr-analyze.sh \
    --base-addr 0x08000000 \
    -s analyze_all.py \
    -o ./fw_analysis \
    firmware.bin

Security Audit

{baseDir}/scripts/angr-analyze.sh -s find_vulns.py -o ./audit target_binary
{baseDir}/scripts/angr-analyze.sh -s export_xrefs.py -o ./audit target_binary
cat ./audit/target_binary_vulns.json | jq '.dangerous_calls'

Troubleshooting

angr Not Installed

pip install angr
# Or in a virtualenv:
python3 -m venv angr-env && source angr-env/bin/activate && pip install angr

Analysis Takes Too Long

{baseDir}/scripts/angr-analyze.sh --timeout 300 -s analyze_all.py binary

Out of Memory on Large Binaries

angr can be memory-intensive. For large binaries:

  • Use individual scripts instead of analyze_all.py
  • Avoid CFGEmulated (CFGFast is used by default and is much lighter)
  • Increase system swap or use a machine with more RAM

Wrong Architecture

angr auto-detects architecture from ELF/PE headers. For raw firmware, specify the base address with --base-addr.

Tips

  1. Start with analyze_all.py — gives a full overview for triage
  2. CFGFast vs CFGEmulated — scripts use CFGFast (fast, incomplete) by default; CFGEmulated is more accurate but much slower
  3. Stripped binaries work — angr recovers functions even without symbols
  4. Symbolic execution needs targets — always specify --find address for symbolic_explore.py
  5. Use jq for JSON — all JSON exports are designed for machine processing
  6. angr's decompiler is experimental — cross-reference with Ghidra for critical analysis

Sponsors

Details

Marketplace
sandbornm/my-claude-skills
Source
angr-analyzer
Category
Community

More from sandbornm/my-claude-skills

All 5