Liberty91LTD/cti-skills / skills/stix-bundle
stix-bundle
VerifiedSTIX 2.1 bundle creation reference. Object types, relationships, and JSON templates for structured threat intelligence sharing. Includes the per-claim mapping of evidence grades (access level and claim support) from /quality-of-information-check to STIX confidence, and the reverse mapping on import.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 14.7 KB |
SKILL.md
View on GitHub---
name: stix-bundle
description: STIX 2.1 bundle creation reference. Object types, relationships, and JSON templates for structured threat intelligence sharing. Includes the per-claim mapping of evidence grades (access level and claim support) from /quality-of-information-check to STIX confidence, and the reverse mapping on import.
user-invocable: false
metadata:
version: 2.0.0
---
# STIX 2.1 Bundle Creation
STIX (Structured Threat Information Expression) 2.1 is the standard for representing and sharing cyber threat intelligence.
## Bundle Structure
```json
{
"type": "bundle",
"id": "bundle--<UUID>",
"objects": [
// Array of STIX Domain Objects (SDOs) and Relationship Objects (SROs)
]
}
```
## STIX Domain Objects (SDOs)
### Indicator
Represents a pattern that can be used to detect suspicious activity.
```json
{
"type": "indicator",
"spec_version": "2.1",
"id": "indicator--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"name": "Malicious IP used by APT28",
"description": "C2 server observed in campaign targeting European government entities",
"pattern": "[ipv4-addr:value = '203.0.113.42']",
"pattern_type": "stix",
"valid_from": "2026-01-15T00:00:00.000Z",
"valid_until": "2026-04-15T00:00:00.000Z",
"kill_chain_phases": [
{
"kill_chain_name": "mitre-attack",
"phase_name": "command-and-control"
}
],
"confidence": 70,
"labels": ["malicious-activity"],
"object_marking_refs": ["marking-definition--613f2e26-407d-48c7-9eca-b8e91df99dc9"]
}
```
### STIX Pattern Syntax
| IOC Type | STIX Pattern |
|----------|-------------|
| IPv4 | `[ipv4-addr:value = '1.2.3.4']` |
| IPv6 | `[ipv6-addr:value = '2001:db8::1']` |
| Domain | `[domain-name:value = 'evil.com']` |
| URL | `[url:value = 'https://evil.com/payload']` |
| SHA-256 | `[file:hashes.'SHA-256' = 'abc123...']` |
| MD5 | `[file:hashes.MD5 = 'abc123...']` |
| Email | `[email-addr:value = 'phish@evil.com']` |
| File name | `[file:name = 'malware.exe']` |
| Combined | `[file:hashes.'SHA-256' = 'abc...' AND file:name = 'payload.dll']` |
### Threat Actor
```json
{
"type": "threat-actor",
"spec_version": "2.1",
"id": "threat-actor--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"name": "APT28",
"description": "Russian state-sponsored threat actor also known as Fancy Bear, Sofacy",
"aliases": ["Fancy Bear", "Sofacy", "Pawn Storm", "Forest Blizzard"],
"threat_actor_types": ["nation-state"],
"roles": ["agent"],
"sophistication": "expert",
"resource_level": "government",
"primary_motivation": "political",
"goals": ["espionage", "disruption"]
}
```
### Malware
```json
{
"type": "malware",
"spec_version": "2.1",
"id": "malware--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"name": "SUNBURST",
"description": "Backdoor trojanised into SolarWinds Orion software",
"malware_types": ["backdoor", "trojan"],
"is_family": true,
"kill_chain_phases": [
{"kill_chain_name": "mitre-attack", "phase_name": "initial-access"},
{"kill_chain_name": "mitre-attack", "phase_name": "command-and-control"}
]
}
```
### Attack Pattern (MITRE ATT&CK Technique)
```json
{
"type": "attack-pattern",
"spec_version": "2.1",
"id": "attack-pattern--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"name": "Phishing: Spearphishing Attachment",
"description": "Adversaries send spearphishing emails with malicious attachments",
"external_references": [
{
"source_name": "mitre-attack",
"external_id": "T1566.001",
"url": "https://attack.mitre.org/techniques/T1566/001"
}
],
"kill_chain_phases": [
{"kill_chain_name": "mitre-attack", "phase_name": "initial-access"}
]
}
```
### Campaign
```json
{
"type": "campaign",
"spec_version": "2.1",
"id": "campaign--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"name": "Operation Fancy Storm",
"description": "Targeted espionage campaign against European government entities",
"first_seen": "2025-11-01T00:00:00.000Z",
"last_seen": "2026-01-15T00:00:00.000Z"
}
```
### Identity (Victim/Target)
```json
{
"type": "identity",
"spec_version": "2.1",
"id": "identity--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"name": "European Government Sector",
"identity_class": "class",
"sectors": ["government-national"]
}
```
## Relationship Objects (SROs)
```json
{
"type": "relationship",
"spec_version": "2.1",
"id": "relationship--<UUID>",
"created": "2026-01-15T00:00:00.000Z",
"modified": "2026-01-15T00:00:00.000Z",
"relationship_type": "uses",
"source_ref": "threat-actor--<UUID>",
"target_ref": "malware--<UUID>",
"confidence": 70
}
```
`confidence` on an SRO grades the relationship itself, the claim that the source uses, targets or is attributed to the target. See [Confidence and source grading](#confidence-and-source-grading) for how the value is derived.
### Common Relationship Types
| Source | Relationship | Target |
|--------|-------------|--------|
| threat-actor | uses | malware, attack-pattern, tool |
| threat-actor | targets | identity, vulnerability |
| threat-actor | attributed-to | identity (sponsoring nation) |
| campaign | uses | malware, attack-pattern, tool |
| campaign | targets | identity, vulnerability |
| campaign | attributed-to | threat-actor |
| indicator | indicates | malware, threat-actor, campaign |
| malware | targets | identity, vulnerability |
| malware | uses | attack-pattern |
## Confidence and source grading
Grades are exported **per claim, never per event**. Each evidence item from `/quality-of-information-check` (schema: `skills/quality-of-information-check/references/evidence-item-schema.md`) carries its own `grading.access_level` and `grading.claim_support`, and each lands on the one STIX object that expresses that claim. Do not stamp one value across every object in a bundle, and do not put `confidence` on the bundle itself. A bundle is a container and has no such property.
The evidence grade is not the Admiralty scale. It is two words, access level and claim support, and it is never written or exported as an Admiralty letter or number. Admiralty ratings apply to lookup results and single items; the evidence grade applies to claims from documents; neither is converted into the other.
### Claim support to `confidence`
STIX `confidence` comes from claim support. The values are the pack's own mapping, chosen to sit inside the bands of `/confidence-levels` (High 80 to 100, Moderate 60 to 79, Low 40 to 59). Use these values and no others.
| Claim support | `confidence` on export |
|---|---|
| established | 90 |
| firm | 70 |
| tentative | 50 |
| disputed | 30 |
| unverified | omit the property |
Unverified means the property is left out. It is never written as 0.
### Which object carries it
| Claim | Object that carries `confidence` |
|---|---|
| Any observable on its own (`domain-name`, `ipv4-addr`, `file`, `url`, `email-addr`) | None. SCOs never carry `confidence`. An observable is a fact about the world, not a claim. |
| "This observable is malicious" or "this pattern detects X" | The Indicator. One Indicator per distinct assertion, so the same IP asserted as C2 by one primary and as a scanner by another is two Indicators with separate grades. |
| Relational claims: `indicates`, `based-on`, `uses`, `attributed-to`, `targets`, `exploits` | The Relationship. For "X was seen at Y", the Sighting. |
| Other SDOs (`threat-actor`, `intrusion-set`, `malware`, `campaign`) | Only when the claim is the entity's own existence as a distinct thing, for example a vendor asserting a new cluster. Otherwise leave it off and grade the relationships. |
An attribution is a Relationship (`attributed-to`), so its grade goes there and not on the Threat Actor. A well-known actor object with `confidence: 50` wrongly says the actor's existence is in doubt.
### Access level and the primary source
STIX has no property for how the source knows. On the same object that carries `confidence`:
- `x_liberty91_access_level`: the word, one of `direct`, `limited`, `indirect`, `untraced`, `adversary`.
- `x_liberty91_claim_support`: the word, one of `established`, `firm`, `tentative`, `disputed`, `unverified`.
- An `external_references` entry for the primary source (`source_name`, `url`, and `description` with the report title and date). Cite the primary, not the outlet that relayed it.
- When the grade was read from a confidence value rather than assessed, `x_liberty91_derived_from_confidence: true`.
Both custom properties are still written when claim support is unverified and `confidence` is omitted, for example `adversary` and `unverified` on an uncorroborated actor claim.
**OpenCTI.** Do not write the evidence grade into an Organization's native reliability field. That field is a track-record rating of the source, and the evidence grade belongs to the claim. Put the two words in the labels or the description of the object that expresses the claim. The custom properties still travel in the bundle for other consumers.
### Worked example
Two claims from one vendor report: an observation graded direct, firm and an attribution graded direct, tentative.
```json
[
{
"type": "indicator",
"spec_version": "2.1",
"id": "indicator--<UUID>",
"created": "2026-09-27T00:00:00.000Z",
"modified": "2026-09-27T00:00:00.000Z",
"name": "C2 domain observed in incident response",
"pattern": "[domain-name:value = 'update-check.example.net']",
"pattern_type": "stix",
"valid_from": "2026-09-20T00:00:00.000Z",
"confidence": 70,
"x_liberty91_access_level": "direct",
"x_liberty91_claim_support": "firm",
"external_references": [
{
"source_name": "Example Vendor",
"description": "Primary source. Intrusion report, 2026-09-22",
"url": "https://vendor.example.com/blog/intrusion-report"
}
]
},
{
"type": "relationship",
"spec_version": "2.1",
"id": "relationship--<UUID>",
"created": "2026-09-27T00:00:00.000Z",
"modified": "2026-09-27T00:00:00.000Z",
"relationship_type": "attributed-to",
"source_ref": "campaign--<UUID>",
"target_ref": "threat-actor--<UUID>",
"confidence": 50,
"x_liberty91_access_level": "direct",
"x_liberty91_claim_support": "tentative",
"external_references": [
{
"source_name": "Example Vendor",
"description": "Primary source. Intrusion report, 2026-09-22",
"url": "https://vendor.example.com/blog/intrusion-report"
}
]
}
]
```
The `domain-name` SCO and the `threat-actor` SDO, if included, carry neither `confidence` nor the two grade properties.
### Report objects
A Report groups many claims, so it has no grade of its own. Where a consumer needs a value, write the **weakest link**: the `confidence` of the lowest-graded load-bearing claim (`event_qoi_summary.weakest_link`), with `x_liberty91_confidence_basis: "weakest_link"`. Never an average. If the weakest link is unverified, omit `confidence` and keep the basis property. This value exists for interoperability only. Analysis consumes the per-claim grades.
### Import direction
Reading a bundle from elsewhere, turn `confidence` into claim support with this table.
| `confidence` read on import | Claim support |
|---|---|
| 60 to 100 | firm |
| 40 to 59 | tentative |
| 1 to 39 | disputed |
| absent or 0 | unverified |
An imported value never gives `established`. That level needs counted independent primaries (rubric rule R5), and a number cannot show them. Access level for an imported item is `indirect` unless the primary is resolved. Do not infer it from the producer's name. Set the `derived_from_confidence` flag on every imported item. The number tells you what the producer thought, not how they knew, so `/quality-of-information-check` re-grades the item when a primary is available (rubric rule R11).
An Admiralty rating that arrives on imported data is kept as that party's Admiralty rating and shown as such. It is not converted into an evidence grade.
MISP tagging for the same grades is in `/ioc-export`.
## TLP Marking Definitions (Standard UUIDs)
These are the canonical marking-definition object IDs. Reference them from `object_marking_refs`; **do not include the marking-definition object itself in the bundle** (see below).
```json
"marking-definition--613f2e26-407d-48c7-9eca-b8e91df99dc9" // TLP:CLEAR
"marking-definition--34098fce-860f-48ae-8e50-ebd3cc5e41da" // TLP:GREEN
"marking-definition--f88d31f6-486f-44da-b317-01333bde0b82" // TLP:AMBER
"marking-definition--826578e1-40a3-4b46-a8d7-b76e4fd71d29" // TLP:AMBER+STRICT
"marking-definition--e828b379-4e03-4974-9ac4-e53a884c97c1" // TLP:RED
```
### MISP-compatible bundles — reference TLPs, don't inline them
For bundles that will be imported into MISP via `/lookup-misp` `upload-stix`, **only reference TLP marking-definition UUIDs** in `object_marking_refs`. Don't include a `{"type": "marking-definition", ...}` object in the bundle's `objects` array.
The inline shape that older STIX 2.0 examples show — `definition_type: "tlp"` with `definition: {"tlp": "clear"}` — fails MISP's STIX 2 importer with `"Does not match any TLP Marking definition!"`. MISP resolves the canonical UUIDs from its own catalog; supplying an inline copy triggers the spec-conformance check and the whole bundle is rejected.
✅ **Correct (MISP-compatible)** — reference only, marking resolved by consumer:
```json
{
"type": "bundle",
"id": "bundle--...",
"objects": [
{
"type": "indicator",
"id": "indicator--...",
"object_marking_refs": ["marking-definition--613f2e26-407d-48c7-9eca-b8e91df99dc9"],
"pattern": "[ipv4-addr:value = '203.0.113.42']",
"pattern_type": "stix",
"valid_from": "2026-04-26T00:00:00.000Z"
}
]
}
```
❌ **Don't do this** for MISP-bound bundles — inlined TLP definition trips the importer:
```json
{
"type": "marking-definition",
"id": "marking-definition--613f2e26-407d-48c7-9eca-b8e91df99dc9",
"definition_type": "tlp",
"definition": {"tlp": "clear"}
}
```
If you need a self-contained bundle for non-MISP consumers (offline ingest, archives, partners without a TLP catalog), the inline form is valid STIX 2.1 — keep a separate non-MISP variant rather than tweaking the MISP one.
If you need TLP applied inside MISP after import, do it via a `tag-event` call rather than a STIX marking — see `tools/integrations/misp.md`.
**OpenCTI is not affected** — it is STIX 2.1-native and imports bundles as-is via `/lookup-opencti` `upload-stix`, inline TLP marking-definitions included. Only MISP-bound bundles need the reference-only treatment above.
## Output Location
Write STIX bundles to: `data/stix-bundles/YYYY-MM-DD-<context>.json`
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 472k | Updated | Claude CodeCodex | Verified | Install |
| 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: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |