Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add meltedinhex/analyst-ai-pack --skill detecting-sandbox-evasion-behaviorgit clone --depth 1 https://github.com/meltedinhex/analyst-ai-packWrote 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/meltedinhex/analyst-ai-pack/detecting-sandbox-evasion-behavior)<a href="https://agentmods.dev/skills/meltedinhex/analyst-ai-pack/detecting-sandbox-evasion-behavior"><img src="https://agentmods.dev/badge/skills/meltedinhex/analyst-ai-pack/detecting-sandbox-evasion-behavior/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/meltedinhex/analyst-ai-pack/detecting-sandbox-evasion-behavior"><img src="https://agentmods.dev/badge/skills/meltedinhex/analyst-ai-pack/detecting-sandbox-evasion-behavior.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00079 | $0.00689 |
| Opus 5 | $0.00039 | $0.00345 |
| Sonnet 5 | $0.00016 | $0.00138 |
| Haiku 4.5 | $0.00008 | $0.00069 |
Grade A, and why
detecting-sandbox-evasion-behavior 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.
What it actually says
Detecting Sandbox Evasion Behavior
When to Use
- A sample ran inertly in a sandbox and you suspect evasion.
- You want to enumerate anti-VM, anti-debug, timing, and environment-fingerprinting checks from static strings/imports and/or an API trace.
Do not use this to conclude a sample is benign — evasion is itself a malicious signal, and absence of detected checks does not prove there are none.
Prerequisites
- The sample (read inertly) and optionally a JSON API trace.
Safety & Handling
- Read bytes statically; never execute the sample to "see if it evades."
Workflow
Step 1: Scan for evasion indicators
python scripts/analyst.py scan sample.bin
Searches for VM/sandbox artifact strings (VMware, VBOX, vmtoolsd, sbiedll, common
sandbox usernames/hostnames), anti-debug APIs (IsDebuggerPresent, CheckRemoteDebuggerPresent,
NtQueryInformationProcess), timing stalls (Sleep, GetTickCount, rdtsc), and CPUID checks.
Step 2: Categorize techniques
Group hits into anti-VM, anti-debug, anti-sandbox, and timing/stalling categories.
Step 3: Recommend bypass and report
Suggest analysis adjustments (patch sleeps, hardened VM, hooking) and map findings to ATT&CK.
Validation
- Detected indicators are grouped by evasion category.
- Findings distinguish static-string evidence from API-trace evidence when both are provided.
- Each category maps to an ATT&CK technique/subtechnique.
Pitfalls
- False positives from benign software that also queries the environment — corroborate.
- Strings can be obfuscated; absence of plaintext artifacts is not absence of evasion.
- Confusing a hung sample with deliberate stalling without timing evidence.
References
- See
references/api-reference.mdfor the scanner. - ATT&CK T1497 and T1622 (linked in frontmatter).
What ships with it
3 files 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 · 85 lines · 79 tokens per session scan A a9bd195a46ae
detecting-sandbox-evasion-behavior is a skill published in the GitHub repository meltedinhex/analyst-ai-pack (22 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 79 tokens to every session and 689 once invoked, about $0.0004 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
dynamic-malware-analysis
Execute and analyze malware samples within a highly controlled, instrumented sandbox environment to observe their true behavior, network communications, file system modifications, and registry changes in real-time.
analyzing-golang-malware-with-ghidra
Reverse engineer Go-compiled malware in Ghidra by parsing Go buildinfo and pclntab structures, recovering stripped/obfuscated function names (e.g. via GoResolver), and extracting embedded module/dependency strings and types from Go binaries. Use when analyzing a Go-language malware sample, deobfuscating a…
analyzing-malicious-pdf-with-peepdf
Perform static analysis of malicious PDF documents using peepdf, pdfid, and pdf-parser to extract embedded JavaScript, shellcode, and suspicious objects. Use when triaging a suspicious PDF attachment from a phishing email, analyzing a PDF-based exploit document, or building detection signatures for weaponized PDF…
dynamic-malware-analysis
Execute and analyze malware samples within a highly controlled, instrumented sandbox environment to observe their true behavior, network communications, file system modifications, and registry changes in real-time.
analyzing-malicious-pdf-with-peepdf
A Chinese-language skill for examining suspicious PDF files with peepdf, pdfid, and pdf-parser. It is intended for static malware analysis, which studies a file without running it.
analyzing-malicious-pdf-with-peepdf
Perform static analysis of malicious PDF documents using peepdf, pdfid, and pdf-parser to extract embedded JavaScript, shellcode, and suspicious objects.