meltedinhex/analyst-ai-pack / skills/performing-dynamic-analysis-in-a-sandbox
performing-dynamic-analysis-in-a-sandbox
VerifiedRuns a sample in an instrumented sandbox to observe behavior: process tree, file and registry changes, network activity, and persistence, then summarizes the behavioral report into capabilities and IOCs. Activates for requests to detonate a sample, do dynamic or behavioral analysis, or interpret sandbox output.
Performing Dynamic Analysis in a Sandbox
When to Use
- Static analysis is blocked by packing/obfuscation and you need to observe runtime behavior.
- You want a behavioral picture: spawned processes, dropped files, registry edits, and C2.
- You are confirming capabilities hypothesized from static analysis.
Do not use dynamic analysis as the only method — evasive samples may sleep, detect the sandbox, or require arguments. Pair it with static/RE work.
Prerequisites
- An isolated lab with a victim VM and simulated internet (see the lab-setup skill).
- A sandbox stack (CAPE/Cuckoo) or manual instrumentation: Process Monitor, Process Explorer, Regshot, and a packet capture on the services guest.
- A clean base snapshot to revert to.
Safety & Handling
- Detonate only inside the isolated victim VM; revert the snapshot after each run.
- Route all network through the simulated-internet guest; never allow live egress.
- Defang any captured URLs/IPs before they leave the lab.
Workflow
Step 1: Prepare instrumentation
Start Process Monitor (filtered to the target), Regshot baseline, and packet capture on the services guest. Snapshot the victim as clean.
Step 2: Detonate with the right context
Many samples need a parent (Office, rundll32), an export (rundll32 dll,Export), or arguments.
Match the original delivery context or the sample stalls.
Step 3: Observe for the full behavior window
Watch process creation, file drops, registry Run keys/services, scheduled tasks, and network beacons. Give it several minutes; some samples sleep first.
Step 4: Summarize the report
Feed the sandbox JSON (or your collected logs) to the summarizer to group events into capabilities and IOCs:
python scripts/analyst.py summarize report.json
Step 5: Diff the system state
Compare Regshot/file baselines pre- and post-run to capture persistence and dropped artifacts.
Validation
- The observed process tree and network match the static-analysis hypothesis.
- Persistence mechanisms found dynamically are confirmed in registry/task artifacts.
- Re-running from the clean snapshot reproduces the core behavior.
Pitfalls
- Detonating without the required parent/args, so the sample exits early ("sandbox evasion" that is really a context problem).
- Too-short observation windows that miss sleeping or staged payloads.
- Trusting a single sandbox verdict; tune and corroborate.
- Allowing real network egress and contaminating live infrastructure.
References
- See
references/api-reference.mdfor the report summarizer. - CAPE Sandbox and Process Monitor (linked in frontmatter).
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 |