Skip to content

meltedinhex/analyst-ai-pack / skills/hunting-persistence-mechanisms-on-windows

hunting-persistence-mechanisms-on-windows

Verified

Hunts for Windows persistence across autostart locations: Run keys, services, scheduled tasks, WMI event subscriptions, and startup folders, baselining them to surface anomalous or recently added entries. Activates for requests to hunt persistence, audit autostart extensibility points, or find malware autoruns on Windows.

Hunting Persistence Mechanisms on Windows

When to Use

  • You are hunting for how an adversary maintains a foothold across reboots.
  • You have autostart inventories (Autoruns exports, registry/scheduled-task/WMI dumps) from one or many hosts and want to surface anomalies.
  • You want to baseline a fleet and flag entries that deviate or were recently added.

Do not use a single host's autoruns list as ground truth — many legitimate entries look unusual; comparison against a fleet baseline is what makes anomalies meaningful.

Prerequisites

  • Autostart data per host: Run/RunOnce keys, services, scheduled tasks, WMI event subscriptions, startup folders, and (ideally) signing status.
  • A baseline (golden image or fleet majority) to diff against.

Workflow

Step 1: Inventory the ASEPs

Collect the common Autostart Extensibility Points across hosts: Run keys (T1547.001), services (T1543.003), scheduled tasks (T1053.005), and WMI subscriptions (T1546.003).

Step 2: Baseline and diff

Compare each host's entries to the fleet baseline; entries appearing on few hosts, unsigned, or pointing to user-writable/temp paths bubble up.

python scripts/analyst.py hunt autostart.json --baseline baseline.json

Step 3: Score anomalies

Raise priority for: unsigned binaries, LOLBins launched with encoded args, paths in %TEMP%/%APPDATA%, recently created tasks, and WMI consumers running scripts.

Step 4: Triage candidates

For each high-score entry, pull the target binary's signing/hash, the creating user/time, and correlate with process and network telemetry.

Step 5: Confirm and respond

Validate suspicious entries (sandbox/RE the binary), escalate confirmed persistence to IR, and turn the pattern into a detection.

Validation

  • Anomalies are ranked against a baseline, not flagged from one host in isolation.
  • High-score entries map to concrete artifacts (path, hash, signer) for triage.
  • Confirmed persistence ties back to a known intrusion timeline.

Pitfalls

  • Treating every unsigned entry as malicious; many legitimate tools are unsigned.
  • Checking only Run keys and missing services/tasks/WMI subscriptions.
  • No timestamps, so you cannot distinguish long-standing from newly planted entries.

References

rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall