Borrowing it
Nothing to install: this file belongs to N0zoM1z0/th08. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/N0zoM1z0/th08/main/.agents/skills/th08-re/SKILL.mdgit clone --depth 1 https://github.com/N0zoM1z0/th08Wrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/n0zom1z0/th08/th08-re)<a href="https://agentmods.dev/skills/n0zom1z0/th08/th08-re"><img src="https://agentmods.dev/badge/skills/n0zom1z0/th08/th08-re/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/n0zom1z0/th08/th08-re"><img src="https://agentmods.dev/badge/skills/n0zom1z0/th08/th08-re.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00069 | $0.01184 |
| Opus 5 | $0.00034 | $0.00592 |
| Sonnet 5 | $0.00014 | $0.00237 |
| Haiku 4.5 | $0.00007 | $0.00118 |
Grade A, and why
th08-re scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TH08 reconstruction
Work only against the target in config/target.toml: Japanese 1.00d, SHA-256
330fbdbf58a710829d65277b4f312cfbb38d5448b3df523e79350b879213d924,
image base 0x00400000, entry 0x004A619E, and .text
0x00402000–0x004B3B77.
Preflight
-
Read
AGENTS.md,docs/ARCHITECTURE.md,docs/RE_WORKFLOW.md, the relevantconfig/mapping.csvandconfig/reccmp-functions.csvrows, anddocs/RE_HANDOFF.md. -
Run:
python3 scripts/verify-target.py python3 scripts/validate-tracking.py --require-target -
Fail closed on IDA until its active TH08 database passes
docs/IDA_MCP.md. The canonical SHA belongs toresources/th08.exe; if IDA reports a different loader/database hash, establish mapped-image identity with the required metadata and multi-point read-only byte samples. Use verified-targetobjdumpor the dedicated headless Ghidra import until then. Never use IDA semantic evidence as a matching claim. -
Confirm
config/claims.csvis header-only and keep one bounded writable scope. Reconcile mapping/Ghidra boundaries against complete target control flow.
Recover and implement
- Record facts as Observed, Inferred, or Unknown. Target bytes, imports, relocations, and directly decoded accesses are observations.
- Inspect the full function plus callers, callees, adjacent code, strings,
globals, access widths, stack cleanup, and saved registers.
Prioritize coherent high-ROI clusters in one translation unit: a sub-0x20
getter or predicate may unlock a nearby callback's
REL32, and a neighboringCutChainoften shares the same object profile and chain globals. Investigate getter, callback, thenCutChain; adjacency is a prioritization signal, not proof of boundaries, ownership, or semantics. - Use the inherited TH08 source as a hypothesis when it is stubbed or incomplete. Use TH06 and TH07 only as semantic or compiler corroboration.
- Recover calling convention, return behavior, field offsets, side effects,
and exceptional paths before shaping source.
For a large switch dispatcher that is not exact yet, audit source presence
with three independent surfaces: enumerate every primary and nested jump-
table slot including shared/default handlers; compare the target direct-call
destination multiset with the built function's
REL32multiset; and reconcile returns, fallthroughs, field offsets, and access widths. This can justifyimplemented.csv, nevermatches.csv. RunEcl is already exact. Its opcode audit is retained atscripts/analysis/historical/runecl-audit-dispatch.pyonly for reproducing the completed investigation; do not select it as new work from stale notes. - Implement the smallest coherent function in its existing module. Do not
manufacture behavior, paste decompiler output, patch target bytes, or use
assembly/byte arrays/padding to force code shape.
If a semantically exact leaf differs only by compiler profile or epilogue,
inspect its address cluster and callers before reshaping the body: a member
defined for one subsystem may have been emitted in another subsystem's
translation unit. Preserve the class owner while testing the target-proven
TU/profile, as with
GameManager::SetYoukaiGaugeinPlayer.obj. - Treat shared layouts, canonical mapping names,
config/implemented.csv, andconfig/match-units.tomlas high-risk changes. Re-run every affected exact unit before committing them. Before publishing aggregate totals after a shared header/layout/compiler/object-graph change, runpython3 scripts/analysis/verify-exact-units.py --all; its cold build is part of the evidence, and--reuse-buildis not sufficient. - Use
$th08-matchingfor the focused VC7 build and strict comparison.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 95 lines · 69 tokens per session scan A d9ef9827c592
th08-re is a skill published in the GitHub repository N0zoM1z0/th08 (92 stars, last pushed today), licensed MIT. It adds 69 tokens to every session and 1,184 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
idapython
IDA Pro Python scripting for reverse engineering. Use when writing IDAPython scripts, analyzing binaries, working with IDA's API for disassembly, decompilation (Hex-Rays), type systems, cross-references, functions, segments, or any IDA database manipulation. Covers ida modules (50+), idautils iterators, and common…
pwn-aws-gamelift
Drive PWN::AWS::GameLift from pwneval.
bg3-steam-launcher
Launch Baldur's Gate 3 through Steam on macOS and load saved games using macos-automator and peekaboo MCP servers. Designed for testing bg3se-macos (Script Extender) development. Use when: (1) launching BG3 from Steam, (2) loading a BG3 saved game, (3) testing SE mod injection, (4) user asks to "start BG3", "load my…
reverse-engineering-binaries-with-ghidra
Uses Ghidra to disassemble and decompile a binary, navigate to key routines via imports and strings, annotate decompiled code, and run headless scripts to automate extraction of C2, crypto, and config. Activates for requests to reverse engineer with Ghidra, decompile a binary, or script Ghidra headless analysis.
analyzing-android-dex-malware
Reverses Android malware: unpacking APKs, decompiling DEX bytecode to readable Java, auditing the manifest for abused permissions and components, and locating dynamically loaded or native payloads. Activates for requests to analyze an APK, decompile DEX, or investigate a suspicious Android app.
analyzing-dotnet-malware-internals
Reverses .NET/managed malware: decompiling MSIL back to C#, defeating common .NET protectors and string encryptors, and tracing reflection-based loaders to recover the real payload. Activates for requests to analyze a .NET sample, decompile MSIL, or unpack a managed loader.