lookup-opencti

Two-way bridge to an OpenCTI threat-intelligence platform. OpenCTI is a STIX 2.1-native knowledge base: everything in it — indicators, observables, intrusion sets, malware, reports, campaigns — is a STIX entity connected by relationships. Like /lookup-misp, this skill also writes: it can create indicators and observables, label and TLP-mark entities, relate them, and import STIX 2.1 bundles. Treat write commands as deliberate actions; don't use them in autonomous enrichment loops. The delete command is destructive — always confirm with the user before invoking it.

When to invoke

Read:

• An indicator was found by another lookup (VT, OTX, …) — check whether your OpenCTI instance already knows it (lookup)
• An analyst wants the entities, reports, or campaigns your platform holds on an actor, malware family, or CVE (search, list, get)
• You need recent high-score indicators, this quarter's incidents, or everything carrying a given label (list with filters)
• You want to verify connector health or instance version (connectors, version)

Write:

• A finished investigation produced indicators that belong in your knowledge base
• A STIX 2.1 bundle written via /stix-bundle should be imported as entities
• An existing entity needs labels, a TLP marking, a score update, or a relationship to another entity

Do NOT invoke for:

• Bulk reputation enrichment (use VT/OTX/RL first; only push the vetted subset into OpenCTI)
• Scanning or pivoting on raw internet infrastructure — OpenCTI is a curated dataset, not a scanning service. Use Shodan/Censys for that.
• Community sharing — that's MISP's job (/lookup-misp). OpenCTI is your internal knowledge graph; MISP is the exchange layer. Many teams run both, synced by a connector.

How to invoke

Single Python CLI (stdlib only — no install). Single endpoint: everything is GraphQL against $OPENCTI_URL/graphql with Authorization: Bearer $OPENCTI_TOKEN.

Read operations

# Connectivity + version check
python3 tools/clis/opencti.py version

# Is this IOC known? Checks observables AND indicators in one call
python3 tools/clis/opencti.py lookup 1.2.3.4
python3 tools/clis/opencti.py lookup evil.example.com --limit 10

# Global full-text search across all entity types (actors, malware, reports, ...)
python3 tools/clis/opencti.py search "APT28" --limit 10

# List/filter one entity type. Types: indicators, observables, reports, incidents,
# campaigns, intrusion-sets, malware, attack-patterns, vulnerabilities, sightings
python3 tools/clis/opencti.py list indicators --score-gte 80 --created-after 2026-06-01T00:00:00Z
python3 tools/clis/opencti.py list reports --label apt28 --limit 10
python3 tools/clis/opencti.py list intrusion-sets --order-by modified --order desc

# Full detail + first 25 relationships for one entity (id from lookup/search/list)
python3 tools/clis/opencti.py get <internal-id>

# Connector health (which enrichment/import connectors are registered + active)
python3 tools/clis/opencti.py connectors

list paginates: when the response has has_next_page: true, pass the returned end_cursor back via --after for the next page.

Write operations

# Create an indicator (auto-generates the STIX pattern; creates the underlying
# observable too by default so raw-value lookups will find it)
python3 tools/clis/opencti.py create-indicator \
  --value 185.220.101.45 --type ip \
  --score 85 --description 'C2 observed 2026-07-01' \
  --labels apt28,c2

# Explicit STIX pattern (for anything the --type shorthand doesn't cover)
python3 tools/clis/opencti.py create-indicator \
  --pattern "[file:hashes.'SHA-256' = '<hash>' AND file:name = 'dropper.exe']" \
  --observable-type StixFile --name 'APT28 dropper'

# Create a bare observable (a fact, no detection semantics)
python3 tools/clis/opencti.py create-observable --value 185.220.101.45 --type ip --labels apt28

# Label + TLP-mark an existing entity (labels are auto-created if missing)
python3 tools/clis/opencti.py add-label <id> apt28 c2
python3 tools/clis/opencti.py add-marking <id> TLP:AMBER

# Update a field (name | description | score | confidence)
python3 tools/clis/opencti.py update <id> --field score --value 95

# Relate two entities with a STIX relationship (indicates, uses, targets, ...)
python3 tools/clis/opencti.py create-relationship --from <indicator-id> --to <malware-id> --type indicates

# Import a STIX 2.1 bundle (upload + trigger import job)
python3 tools/clis/opencti.py upload-stix data/stix-bundles/2026-07-01-apt28-c2.json

# Delete an entity — DESTRUCTIVE. Confirm with the user first.
python3 tools/clis/opencti.py delete <id>

All commands accept --dry-run (preview the GraphQL request without sending) and --insecure (skip TLS verification, common on internal deployments).

The CLI exits 2 if OPENCTI_URL or OPENCTI_TOKEN is unset (when not in dry-run). Report missing credentials; do not fabricate.

Indicators vs. observables

OpenCTI separates the two, and it matters for both reads and writes:

• Observable — a fact: "this IP exists in our data." No detection semantics, no validity window.
• Indicator — an assertion: "this pattern is malicious," with a STIX pattern, score, and valid-from/until window.

lookup checks both and reports which matched (known_as). When writing: use create-indicator for anything you'd want detections or blocklists built from, and let it create the companion observable (the default — don't pass --no-observable unless you have a reason; without the observable, raw value-searches won't find the IOC). Use create-observable for context-only sightings. Deleting an indicator does not cascade to its observable, and vice versa.

STIX 2.1 round-tripping

OpenCTI is STIX-native, so bundles from /stix-bundle import without the marking-definition surgery MISP needs:

1. Run an investigation (e.g. /ip-investigation)
2. Build a STIX bundle from the findings (/stix-bundle)
3. Apply TLP + Admiralty (/apply-tlp, /score-source)
4. python3 tools/clis/opencti.py upload-stix <bundle.json>
5. Review in OpenCTI → add context → connect to actors/campaigns

Import semantics: by default the CLI triggers the import with bypassValidation, which imports the bundle directly. With --no-bypass (or if your token lacks the bypass capability) the bundle lands in the analyst workbench (Data → Import) awaiting human validation — uploaded but not yet in the knowledge base. The CLI's response note tells you which happened; relay it. Import is asynchronous — poll with lookup/search; small bundles take seconds.

Response format

All commands return JSON:

source: opencti
operation: lookup | search | list | get | create-indicator | upload-stix | ...
query_time: <ISO8601>
# lookup:
known: true|false
known_as: [observable, indicator]
observables: [...]
indicators: [...]
# list:
total_matches: <n>
has_next_page: true|false
end_cursor: <cursor for --after>
results: [...]
# get:
entity: {...}
relationships: [...]

Source reliability (Admiralty default)

Default rating for downstream /score-source: B2 (usually reliable, probably true). OpenCTI is an aggregation platform — the rating belongs to the feed or author behind the entity, not the platform. Check createdBy and the entity's labels/markings:

• Your own analysts' entities (self-hosted instance) → A1 — primary source
• Reputable vendor connector (Mandiant, CrowdStrike, Recorded Future) → A2 / B2
• Open feeds (AlienVault, abuse.ch, MISP feed connectors) → B3 / C3
• confidence field on the entity (0-100) maps directly onto /confidence-levels bands

Operational notes

• GraphQL errors arrive as HTTP 200 with an errors array — the CLI surfaces the first message and exits 1. AUTH_REQUIRED means a bad or missing token.
• Scores drive automation. x_opencti_score feeds downstream decay and detection-export rules on many deployments. Don't inflate scores on write; 50 is neutral.
• Labels are shared vocabulary. add-label and --labels auto-create missing labels — check list indicators --label <name> for the existing convention before inventing a new one.
• Deletes are silent and permanent. No recycle bin. Confirm with the user, and remember the indicator/observable non-cascade.
• pycti exists — for complex workflows (bulk imports with author identities, custom connectors, streaming) the official pycti library is the right tool. This CLI covers the 80% case for query + write, stdlib-only.

Related skills

• /stix-bundle — write STIX 2.1 bundles that this CLI's upload-stix consumes (OpenCTI accepts them as-is; no MISP-style TLP workaround needed)
• /lookup-misp — the sharing-platform sibling; OpenCTI = internal knowledge graph, MISP = community exchange
• /lookup-virustotal, /lookup-otx — community reputation lookups; chain before pushing into OpenCTI
• /score-source — apply Admiralty rating before publishing
• /apply-tlp — decide TLP before add-marking
• /ip-investigation, /domain-investigation, /hash-investigation, /url-investigation — produce findings that OpenCTI write commands consume
• /threat-actor-profiling, /campaign-tracking — consume search/list/get output for actor and campaign context

See also

• Integration setup: tools/integrations/opencti.md
• Python CLI source: tools/clis/opencti.py
• OpenCTI project: https://filigran.io/solutions/open-cti/
• Documentation: https://docs.opencti.io/
• pycti (full SDK, when this CLI isn't enough): https://github.com/OpenCTI-Platform/client-python
• GraphQL playground: $OPENCTI_URL/public/graphql (when enabled by the admin)