agentleFS
Liberty91LTDLiberty91LTD/cti-skills/skills/hash-investigation/SKILL.md

hash-investigation

Skill35 starsChanged 47 days agoMITView on GitHub

Use when a user asks to check, identify, or characterize a file hash (MD5, SHA-1, SHA-256). Chains VirusTotal and OTX, optionally triggers /malware-analysis for deeper behavioral review. Returns detection signals, malware family, behavioral tags, and pivot candidates (communicating IPs, dropped files). Invoked by /cti-orchestrator when the target is a hash.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If VirusTotal detection ratio is 0/Y with Y>30, AND OTX has no pulses, AND the file has standard legit characteristics (signed by major …

    line 62
  2. Very specific:

    If detection ratio is high and malware family is known, the lookup may be sufficient.

    line 115

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

hash-investigation

Investigate a file hash end-to-end. Chain lookups, surface malware family attribution, prioritize infrastructure pivots, and decide whether deeper malware analysis is warranted.

This skill invokes: /lookup-liberty91 (first-party — run first), /lookup-virustotal, /lookup-otx, optionally /lookup-reversinglabs (vendor-authoritative classification + MITRE ATT&CK + sandbox), optionally /lookup-crowdstrike (Falcon Intel verdict + actor/malware attribution), optionally /lookup-misp and/or /lookup-opencti (internal correlation), optionally /lookup-sentinel (exposure check in your own telemetry), optionally /malware-analysis, then /source-assessment, /tlp-guide, /confidence-levels.

When to invoke

  • User asks to check a hash, identify malware, or enrich a sample
  • Another skill surfaces a hash during investigation (e.g., /ip-investigation found communicating files)
  • User asks "what is this malware" with a hash as input

Inputs

indicator: <MD5 | SHA-1 | SHA-256 hash>
context: <optional — where did this hash come from?>
tlp_ceiling: <optional, default AMBER>
request_malware_analysis: <optional bool — trigger /malware-analysis after lookup>

Procedure

1. Validate

  • Confirm the hash is a valid length (MD5=32, SHA-1=40, SHA-256=64 hex chars)
  • Reject if not; route to correct skill if it looks like a different indicator type

2. Parallel lookups

Always run (free / commonly available):

/lookup-virustotal hash <value>    # detection stats, family, behavior tags
/lookup-otx hash <value>           # community pulses, campaign attribution

Add to the same parallel batch if credentials are configured — don't demote these to "later, maybe":

/lookup-reversinglabs hash <value> --av-scanners --ticloud   # if $REVERSINGLABS_USER is set. Vendor-authoritative verdict + threat name + AV detection ratio. RL is the strongest single-source verdict on a hash; run it whenever available, not just when VT signal is thin.
/lookup-crowdstrike indicator <value>                        # if $CROWDSTRIKE_CLIENT_ID is set. Falcon Intel verdict — malicious confidence + the threat actors and malware families CrowdStrike links to this hash + report refs. One API call.
/lookup-liberty91 ioc-lookup <value>                         # if $LIBERTY91_API_KEY is set. First-party — run first: does your own platform already hold this hash, with what verdict, confidence and TLP?
/lookup-misp search-attributes --value <value>               # if $MISP_URL + $MISP_KEY are set. Internal correlation against your own catalogued events.
/lookup-opencti lookup <value>                               # if $OPENCTI_URL + $OPENCTI_TOKEN are set. Internal correlation against your OpenCTI knowledge base (observables + indicators in one call).

Escalation lookup (run sequentially after the parallel batch only if the call below is warranted):

/lookup-reversinglabs report <value> --detailed              # when you need MITRE ATT&CK mapping, sandbox detonation output, or networkthreatintelligence (C2 indicators extracted from the sample). Heavier than `hash` — use when classification alone is not enough.

3. Check if known-benign

If VirusTotal detection ratio is 0/Y with Y>30, AND OTX has no pulses, AND the file has standard legit characteristics (signed by major vendor, common packaging), treat as likely-benign. Stop and summarize. Don't spend analyst time.

4. Consolidate

indicator: <hash>
type: hash
hash_type: md5 | sha1 | sha256
consolidated_verdict: malicious | suspicious | clean | unknown
detection_stats:
  vt_ratio: <X/Y>
  first_seen: <date>
  last_analysis: <date>
  names: [<aliases>]
classification:
  malware_family: <popular_threat_classification from VT>
  tags: [<from VT>]
  attributed_actor: <if OTX pulses mention an actor>
file_details:
  type: <PE, ELF, Office, script, etc.>
  size: <bytes>
  packers: [<if detected>]
  signed: <boolean>
  signer: <if signed>
behavior_summary:
  sandbox_verdicts: <from VT>
  notable_behaviors: [<from VT tags>]
  mitre_techniques: [<if inferrable>]
community_context:
  otx_pulse_count: <n>
  campaign_references: [<...>]
pivot_candidates:
  - type: ip
    value: <c2 ip from communicating>
    reason: c2_comm
  - type: domain
    value: <c2 domain>
    reason: c2_comm
  - type: hash
    value: <parent or dropped>
    reason: child_parent

5. Prioritize pivots

  1. C2 IPs / domains from communicating_files / contacted_domains — active infrastructure
  2. Parent hashes (droppers / packers) — same campaign lineage
  3. Dropped hashes (payloads) — behavioral family
  4. YARA rule matches (if VT exposes them) — strong family clustering
  5. Command-line strings — targeting / configuration hints

6. Decide on deeper analysis

If detection ratio is high and malware family is known, the lookup may be sufficient. If the hash is novel (low VT detection, recent first_seen, no known family), invoke /malware-analysis for a structural and behavioral review.

Also invoke /malware-analysis if:

  • User explicitly asked for analysis (request_malware_analysis: true)
  • This is part of a targeted incident response, not routine IOC triage

6b. Exposure check — was it seen in OUR environment? (if Sentinel is configured)

If $SENTINEL_WORKSPACE_ID + app credentials are set and the verdict is malicious or suspicious, chain /lookup-sentinel to sweep the hash across the workspace's available tables (DeviceFileEvents, DeviceProcessEvents, DeviceImageLoadEvents, EmailAttachmentInfo, … — that skill discovers what exists before querying). Report hits with first/last seen and affected devices; report a miss as "not observed in collected telemetry over <window>", never "not compromised".

7. Apply rigor

Chain:

  • /source-assessment
  • /tlp-guide
  • /confidence-levels

8. Return

title: "Hash Investigation: <hash>"
tlp: AMBER
confidence: <0-100>
summary: <2-3 sentences with malware family if known>
verdict: <...>
detection_stats: <...>
classification: <...>
file_details: <...>
behavior_summary: <...>
community_context: <...>
pivot_candidates: <top 3-5>
source_ratings: <...>
investigation_date: <ISO8601>
recommended_next_steps:
  - <e.g., "run /malware-analysis", "pivot on C2 IP via /ip-investigation">

Special cases

  • Hash not found in VT — doesn't mean clean. Could be very new, rare, or deleted. Check OTX; escalate to /malware-analysis if user has the sample.
  • VT detection very low (1-2 engines) — often false positive or generic heuristic. Downgrade source rating; don't alarm the user without corroboration.
  • Signed by legit vendor but flagged — possible supply-chain scenario. Flag for special review.

Related skills

  • /ip-investigation, /domain-investigation, /url-investigation
  • /malware-analysis — for novel or targeted samples
  • /yara-writing — if the sample should produce a detection rule
  • /indicator-pivoting — multi-hop IOC graphing
  • /cti-orchestrator

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.