hypnguyen1209/offensive-claude / skills/reverse-engineering
reverse-engineering
VerifiedUse when reverse-engineering a binary or firmware — static triage + decompilation (Ghidra/IDA/Binary Ninja), dynamic instrumentation (GDB/Frida 17/angr), anti-reversing & packer bypass, OLLVM/VM deobfuscation, UEFI/BIOS RE & Secure Boot research, patch-diffing for n-days
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.4 KB |
| references/anti-reversing-bypass.md | markdown | 7.7 KB |
| references/coverage-reachability.md | markdown | 3.2 KB |
| references/deobfuscation.md | markdown | 8.6 KB |
| references/dynamic-instrumentation.md | markdown | 7.9 KB |
| references/firmware-uefi.md | markdown | 8.6 KB |
| references/patch-diffing-protocol.md | markdown | 8.4 KB |
| references/rr-time-travel.md | markdown | 3.1 KB |
| references/static-triage-decompilation.md | markdown | 8.2 KB |
| scripts/antidebug_unhook.py | script | 5.9 KB |
| scripts/cve_diff.py | script | 9.6 KB |
| scripts/deflatten_triton.py | script | 5.9 KB |
| scripts/frida_universal.js | script | 8.3 KB |
| scripts/ghidra/DecompileToC.java | other | 1.8 KB |
| scripts/patchdiff_fetch.py | script | 5.3 KB |
| scripts/proto_infer.py | script | 7.1 KB |
| scripts/rr_root_cause.sh | script | 3.3 KB |
| scripts/string_decrypt_emu.py | script | 6.0 KB |
| scripts/triage.py | script | 9.2 KB |
| scripts/uefi_triage.py | script | 6.2 KB |
references/coverage-reachability.md
View on GitHub# Coverage / Trace Reachability — proving a native bug's line actually ran
A static crash hypothesis ("`strcpy` here overflows") is a *claim* until you show the vulnerable line
executed with your input. This is the produce-side recipe for the reachability artifacts the harness
(`validate_findings.py`) requires before a native memory-corruption finding can be `[CONFIRMED]` — see
`skills/references/finding-evidence-standards.md`. Run these in the binary-analysis devcontainer
(`.devcontainer/`); they need a Linux toolchain (gcc/clang + gcov).
## Option A — gcov line-hit (source available)
Build with coverage, run the PoC/witness, generate gcov, assert the vulnerable line is not `#####`.
```bash
# 1. instrument
gcc --coverage -O0 -g -o target_cov target.c # or: clang -fprofile-instr-generate -fcoverage-mapping
# 2. drive the vulnerable path with your input (the crash witness)
./target_cov < poc_input # may crash; .gcda is flushed on normal exit,
# so wrap the sink to return cleanly OR use
# __gcov_dump() / llvm profile before the crash
# 3. emit gcov and locate the vulnerable line
gcov -b target.c # writes target.c.gcov
sed -n '40,45p' target.c.gcov # eyeball: the vuln line must show a count, not '#####'
```
`target.c.gcov` line format is `<count>:<lineno>:<source>`. A `#####` count means the line never
executed; a `-` means it is not code. Record the proof:
```json
"proof": {"coverage_proof": {"gcov": "evidence/target.c.gcov", "line": 42}}
```
> Crash-before-flush pitfall: gcov flushes counters at process exit, which a segfault skips. Either
> compile the sink to log-and-return instead of crashing during the coverage run, call
> `__gcov_dump()` immediately before the faulting statement, or run the coverage build under a harness
> that traps the signal and dumps. ASan + `-fprofile-...` can be combined for a clean abort path.
## Option B — function trace (binary-only, no source)
Prove the vulnerable function was reached at runtime with a call trace, then cite the log.
```bash
# rr (deterministic) — record then list the functions on the crashing run
rr record ./target poc_input
rr replay -a 2>/dev/null | grep -m1 parse_header # or use rr_root_cause.sh
# OR -finstrument-functions build that logs each entered function
# OR a Frida/ltrace/uftrace capture naming the function
uftrace record ./target poc_input && uftrace report | grep parse_header
```
```json
"proof": {"trace_proof": {"log": "evidence/trace.txt", "function": "parse_header"}}
```
## What this is / isn't
- It demonstrates **reachability** (the line/function ran), not **exploitability**. Pair it with the
feasibility verdict + empirical mitigation map (`exploit-feasibility.md`) for the full picture.
- The harness check is mechanical and fail-closed: a missing/empty/`#####`/unparseable artifact keeps
the finding `[POSSIBLE]`. Don't hand-edit a `.gcov` — regenerate it.
- For root-cause (not just reachability), use rr time-travel: `rr-time-travel.md` + `scripts/rr_root_cause.sh`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |