agentleFS
THARAGESHWARAN-SATHYAMOORTHYTHARAGESHWARAN-SATHYAMOORTHY/mitre-attack-skill/skills/mitre-attack/SKILL.md

mitre-attack

Skill3 starsChanged yesterdayMITView on GitHub

Use for any MITRE ATT&CK work: map incidents/reports to TTPs, validate IDs, detection coverage and gaps, Navigator layers, group profiling, emulation plans, Sigma/Atomic mapping, version migration.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. mukul975/Threatswarm★ 77same name
    2. Liberty91LTD/cti-skills★ 35same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Verify every ID and name against live data: before output (get, or validate on the finished draft).

    line 13
  2. Distinctive:

    Never invent behavior: Every mapping cites evidence (quote, log line, report page).

    line 14
  3. Distinctive:

    Use the most specific sub-technique the evidence supports.

    line 17
  4. Distinctive:

    Test for each behavior: is the actor hiding (Stealth) or breaking a defense (Defense Impairment)?

    line 138
  5. Very specific:

    Write the table (template in section 9).

    line 150
  6. Distinctive:

    Check the CTID Adversary Emulation Library for an existing plan for that actor.

    line 163
  7. Very specific:

    Use attack layer (never hand-write tactic names).

    line 166
  8. Very specific:

    Every Enterprise "Defense Evasion" label: look up the technique's current tactic (attack get), which is Stealth or Defense Impairment (or …

    line 170
  9. Very specific:

    Follow with: unmapped behaviors and why, assumptions, validation result (attack validate: PASS).

    line 251

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

MITRE ATT&CK skill

Baseline: ATT&CK v20.1 (v20 released 2026-10-13; v20.1 data patch modified 2026-11-05). Facts and the tool below were checked against live STIX data on 2026-11-14. When live data disagrees with this text (a newer ATT&CK release), trust the live data. ATT&CK ships major releases in April and October plus smaller "Agile" releases (groups/software/campaigns) in between (first: v19.2, Aug 2026).

Hard rules

  1. State the ATT&CK version (and domain) in every deliverable. Get it from attack.py version.
  2. Verify every ID and name against live data before output (get, or validate on the finished draft). Never quote IDs from memory.
  3. Never invent behavior. Every mapping cites evidence (quote, log line, report page). No evidence, no mapping; inference is marked Low confidence with the reason.
  4. Redact before output. Strip secrets, API keys, tokens, passwords, hashes of credentials, PII, salary data, and card numbers from any log, ticket, or connector data before it goes into a table, layer comment, or report. Use [REDACTED-<type>].
  5. No weaponization. Atomic Red Team output is test names, GUIDs, platforms, and executor type only. Do not write exploit code, payloads, or step-by-step attack commands. Emulation plans reference test IDs and are for authorized scope only.
  6. Use the most specific sub-technique the evidence supports; otherwise the parent.
  7. Enterprise content that says "Defense Evasion" or "14 tactics" is pre-v19: migrate it (workflow E).

0. Run the tool and load data (always first)

One standard-library-only CLI ships with this skill at scripts/attack.py (in this skill's base directory). Do not rewrite or re-download it; run the bundled file:

python3 -I <skill-dir>/scripts/attack.py update && python3 -I <skill-dir>/scripts/attack.py version

Always run it as python3 -I <skill-dir>/scripts/attack.py (below: attack), where <skill-dir> is this skill's base directory. Data is downloaded to ~/attack-data. Bundles come from raw.githubusercontent.com/mitre-attack/attack-stix-data (<domain>-attack/<domain>-attack.json, pinned releases <domain>-attack-<X.Y>.json, version list in index.json). If GitHub is unreachable, use WebFetch on https://attack.mitre.org/techniques/<ID>/ and say the data came from the website.

Commands

Global options (after the command): -d enterprise|mobile|ics (default auto: enterprise, then mobile, then ics), --json, --limit N (table rows, 0 = all), --full (untruncated descriptions). Default output is markdown tables. Exit code: 0 ok, 1 validation failed, 2 error.

CommandWhat it does
attack update [--force] [--release 19.1]Refresh bundles (skips if < 24h old); --release saves <domain>-<X.Y>.json
attack version / attack statsData version per domain / live object counts
attack get <ID or name>Full detail for any object (T, TA, G, S, C, M, DET, AN, DC, A): tactics, platforms, parent/subs, mitigations, DET -> AN -> DC with log sources, groups/software/campaigns using it, revoked/deprecated status with replacement chain and what it supersedes
attack search "<text>" [--type group] [--all]Ranked search over names, aliases, descriptions (finds "Cozy Bear" -> G0016 APT29)
attack group|software|campaign <ID, name or alias>Profile: techniques with tactic and source (direct or via campaign), software, campaigns, attribution
attack validate <file or -> [--fix out] [--all]Extracts every ID from text/CSV/JSON/Sigma YAML. Flags unknown, revoked (with replacement and its tactic), deprecated, legacy Defense Evasion, tactic-tag mismatch (single Sigma rule), bad sub-technique, parent-vs-sub overlap. --fix writes a copy with revoked IDs replaced
attack layer --techniques <csv or ->Navigator layer from technique_id[,score][,comment] rows (or free text). Revoked IDs are auto-replaced with a warning
attack layer --from <ID...> [--op union|intersection|diff] [--with-software]Layer for a group/software/campaign, or overlap of several (score = how many use the technique)
attack coverage <detections.csv> [--groups G0016,FIN7] [--top 25] [--layer out.json]Rows technique_id,detection_name,status with status none|logs|rule|tested. Per-tactic coverage %, gaps ranked by number of groups using each technique, optional layer
attack sigma <dir> [--csv out.csv] [--layer out.json] [--groups ...]Reads attack.tXXXX tags from Sigma rules, reports tag issues, then the same coverage report
attack diff <old> <new> [--type all]Added/revoked/deprecated/changed objects between bundles; args are files or enterprise@19.1 (auto-download)
attack atomic <technique> [--no-subs] [--legacy]Atomic Red Team test names, platforms, executor, GUID for a technique and its subs (no commands)

Examples:

A="python3 -I <skill-dir>/scripts/attack.py"
$A get T1562                      # revoked -> shows replacement T1685
$A get T0886 --json               # ICS technique, JSON
$A group "Cozy Bear" --limit 0    # full APT29 profile
$A validate draft_report.md       # run on every deliverable before sending
$A layer --from G0016 G0032 --op intersection -o overlap.json
$A coverage detections.csv --groups "APT29,Lazarus Group" --layer coverage.json
$A diff enterprise@19.1 enterprise@20.1

Layers use Navigator layer format 4.5 (Navigator 5.3.2, checked 2026-11-14 in mitre-attack/attack-navigator: layers/spec/v4.5/layerformat.md, nav-app/package.json). The tool writes one entry per tactic the technique belongs to, the right domain, and the ATT&CK major version from the data. Coverage scoring: none 0, logs 1, rule 2, tested 3; a parent inherits its best sub-technique score, a sub never inherits from its parent.

1. ID formats

PrefixObjectSTIX type
TA####Tacticx-mitre-tactic
T#### / T####.###Technique / sub-techniqueattack-pattern
G####Groupintrusion-set
S####Softwaremalware / tool
C####Campaigncampaign
M####Mitigationcourse-of-action
DET####Detection strategyx-mitre-detection-strategy
AN####Analyticx-mitre-analytic
DC####Data componentx-mitre-data-component
A####Asset (ICS)x-mitre-asset

Relationships: uses, mitigates, detects (DET -> technique), subtechnique-of, attributed-to (campaign -> group), revoked-by, targets (ICS technique -> asset). Analytics hang off a detection strategy via x_mitre_analytic_refs; data components via each analytic's x_mitre_log_source_references. Filter out revoked and x_mitre_deprecated unless doing history. ICS techniques are T08xx/T09xx; mobile and enterprise share the T1xxx range, and groups/software can exist in several domains.

2. Current structure (v20.1)

Enterprise: 15 tactics (since v19)

IDTacticShortname
TA0043Reconnaissancereconnaissance
TA0042Resource Developmentresource-development
TA0001Initial Accessinitial-access
TA0002Executionexecution
TA0003Persistencepersistence
TA0004Privilege Escalationprivilege-escalation
TA0005Stealth (was Defense Evasion, same ID)stealth
TA0112Defense Impairment (new in v19)defense-impairment
TA0006Credential Accesscredential-access
TA0007Discoverydiscovery
TA0008Lateral Movementlateral-movement
TA0009Collectioncollection
TA0011Command and Controlcommand-and-control
TA0010Exfiltrationexfiltration
TA0040Impactimpact

Platforms: Windows, macOS, Linux, ESXi, IaaS, Network Devices, PRE, Office Suite, SaaS, Identity Provider, Containers.

Mobile: 12 tactics (Android, iOS)

Initial Access TA0027, Execution TA0041, Persistence TA0028, Privilege Escalation TA0029, Defense Evasion TA0030 (the Stealth split was Enterprise only), Credential Access TA0031, Discovery TA0032, Lateral Movement TA0033, Collection TA0035, Command and Control TA0037, Exfiltration TA0036, Impact TA0034.

ICS: 12 tactics

Initial Access TA0108, Execution TA0104, Persistence TA0110, Privilege Escalation TA0111, Evasion TA0103, Discovery TA0102, Lateral Movement TA0109, Collection TA0100, Command and Control TA0101, Inhibit Response Function TA0107, Impair Process Control TA0106, Impact TA0105. ICS also has 18 assets (A####).

Live counts, v20.1 (active objects; recompute with attack stats)

DomainTechniquesSubsGroupsSoftwareCampaignsMitigationsDETANDC
Enterprise22747917884257447061773106
Mobile78472012931312521329
ICS78261425105310430040

3. Recent changes you must know (all verified in data)

v18 (Oct 2025): detection model rebuilt. Data sources were replaced by Detection Strategies (DET) -> Analytics (AN, platform-specific logic with tunables) -> Data Components (DC, with log source name and channel). One DET per technique.

v19 (2026-04-28): Defense Evasion split (Enterprise only).

  • TA0005 renamed Stealth (hiding: masquerading, obfuscation, hide artifacts, indicator removal).
  • New TA0112 Defense Impairment (breaking defenses), 57 techniques incl. subs. 60 existing techniques had their tactic list changed.
  • 23 new techniques, 17 revoked. Replacement map (revoked-by):
RevokedReplacement
T1562 Impair Defenses, T1562.001, T1562.006T1685 Disable or Modify Tools
T1562.002 / .008 / .011 / .012T1685.001 Windows Event Log / T1685.002 Cloud Log / T1685.003 Modify or Spoof Tool UI / T1685.004 Linux Audit System Log
T1070.001 / T1070.002 (log clearing)T1685.005 Clear Windows Event Logs / T1685.006 Clear Linux or Mac System Logs
T1562.004 / .007 / .013T1686 Disable or Modify System Firewall / T1686.001 Cloud Firewall / T1686.002 Network Device Firewall (T1686.003 Windows Host Firewall is new)
T1562.009 / T1562.010 / T1562.003T1688 Safe Mode Boot / T1689 Downgrade Attack / T1690 Prevent Command History Logging
T1656 Impersonation / T1672 Email SpoofingT1684.001 Impersonation / T1684.002 Email Spoofing (under new T1684 Social Engineering, tactic Stealth)
  • Also new in v19: T1687 Exploitation for Defense Impairment, T1682 Query Public AI Services, T1683 Generate Content (.001 Written, .002 Audio-Visual), T1027.018 Invisible Unicode. T1211 renamed Exploitation for Stealth.
  • Test for each behavior: is the actor hiding (Stealth) or breaking a defense (Defense Impairment)?

v19.2 (2026-08-05): first Agile release (groups/software/campaigns).

v20 (2026-10-13): AI-related techniques. T1699 Direct Prompt Injection and T1698 Indirect Prompt Injection (Execution), T1700 AI-Enabled Exfiltration (Exfiltration), T1608.007 Stage Capabilities: AI Result Poisoning (Resource Development), T1685.007 Disable or Modify Tools: Impair Autonomous Defenses (Defense Impairment), T1696 Anti-Automation Evasion with .001 Automate CAPTCHA Completion and .002 Avoid CAPTCHA (Stealth), T1697 Steal Application Specific Passwords (Credential Access). Also present: T1588.007 Obtain Capabilities: Artificial Intelligence. No technique revocations between v19.1 and v20.1. Use MITRE ATLAS alongside ATT&CK for attacks on AI systems themselves.

4. Workflows

A. Map a report or incident

  1. List each observed behavior as one sentence: actor did X, with Y, on platform Z. Keep the evidence quote and source location. Redact secrets and PII now.
  2. For each behavior pick the tactic by the actor's goal at that step (table in section 5), then find the technique: attack search "<verb + object>", read attack get <ID> description and compare.
  3. Choose the most specific sub-technique the evidence supports. Assign confidence.
  4. Write the table (template in section 9). Run attack validate on the draft; fix every finding.
  5. Optional: attack layer --techniques mapped.csv -o incident.json.

B. Detection coverage and gaps

  1. Build detections.csv (technique_id,detection_name,status). From Sigma: attack sigma rules/ --csv detections.csv. From SIEM/EDR exports: map each rule to technique IDs, then validate.
  2. attack coverage detections.csv --groups <relevant groups> --layer coverage.json.
  3. For each top gap run attack get <ID>: read its DET -> AN -> DC rows and compare the log sources to what is actually collected. Missing log source means status stays none; collected but no rule means logs.
  4. Report with the coverage template. Say plainly that coverage % measures mapped rules, not detection quality.

C. Threat-informed emulation plan

  1. attack group <name> (or campaign). Add attack layer --from <ID> --with-software for the picture including tooling.
  2. Order techniques by tactic flow. For each, attack atomic <ID> and pick tests by platform; note test GUIDs only.
  3. For each step list expected telemetry (DC and log source from attack get) and the detection that should fire.
  4. Check the CTID Adversary Emulation Library for an existing plan for that actor. Use the emulation template.

D. Navigator layers

Use attack layer (never hand-write tactic names). Open at https://mitre-attack.github.io/attack-navigator/ -> Open Existing Layer. Group overlap: --op intersection for shared tradecraft, --op diff for what is unique to the first entity, union score = number of entities using each technique (heat map for prioritization).

E. Migrate pre-v19 or older content

  1. attack validate old.md --fix migrated.md replaces revoked IDs. It does not change tactic labels or technique names: fix those by hand from the findings.
  2. Every Enterprise "Defense Evasion" label: look up the technique's current tactic (attack get), which is Stealth or Defense Impairment (or both removed).
  3. attack diff enterprise@<old> enterprise@<new> lists added, revoked, renamed, and tactic-moved techniques for release notes. Official changelogs: https://attack.mitre.org/docs/changelogs/.
  4. Record old and new ATT&CK versions in the deliverable.

5. Mapping methodology (CISA "Best Practices for MITRE ATT&CK Mapping" style)

Order of work: find the behavior -> research it -> identify the tactic (why) -> identify the technique (how) -> identify the sub-technique -> peer check. Map behaviors, not indicators: an IP address or hash is not a technique.

Actor goal at this stepTactic
Learn about the target before accessReconnaissance
Build or buy infrastructure, accounts, toolsResource Development
Get inInitial Access
Run codeExecution
Keep access across restarts or credential changesPersistence
Get higher permissionsPrivilege Escalation
Avoid being noticedStealth
Disable, break, or blind a security control or logDefense Impairment
Steal credentials or tokensCredential Access
Learn the environment after accessDiscovery
Move to other systemsLateral Movement
Gather target dataCollection
Talk to compromised systemsCommand and Control
Take data outExfiltration
Destroy, encrypt, disrupt, manipulateImpact

Confidence levels:

  • High: direct evidence of the behavior (command line, log event, malware analysis) matches the technique description.
  • Medium: behavior is clear but the sub-technique or tactic is inferred from context.
  • Low: inferred from an outcome or a tool's known capability; state the assumption.

Common mistakes:

  • Mapping a tool's full capability list instead of what was observed in this incident.
  • One technique per sentence by reflex: one behavior can need two mappings (e.g., a scheduled task is Execution and Persistence only if evidence shows both goals), and many sentences need none.
  • Picking the tactic from the technique's first listed tactic instead of the actor's goal.
  • Parent used when a sub fits, or a sub guessed when evidence only supports the parent.
  • Confusing look-alikes: Valid Accounts (using creds) vs Brute Force (getting them); Exploit Public-Facing Application (initial access) vs Exploitation for Privilege Escalation; Ingress Tool Transfer vs Lateral Tool Transfer; Phishing vs Phishing for Information.
  • Using revoked IDs or pre-v19 tactic names. Mixing domains (ICS T08xx in an Enterprise layer).
  • Treating mitigations or CVEs as techniques.

6. Detection engineering

  • Start from attack get <technique>: each analytic row gives platform, data components, log source names (e.g., WinEventLog:Sysmon, auditd:SYSCALL, AWS:CloudTrail), and a logic summary. attack get AN#### adds log channels (event codes) and tunable fields. attack get DC#### lists every technique detectable from that data component: use it to justify onboarding a log source.
  • Analytics are behavioral descriptions, not ready rules. Translate to your SIEM, tune the listed mutable fields, test with an Atomic test, then mark tested.
  • Tag rules at the most specific level; one rule may carry several techniques.
  • Sigma tags: lowercase attack.<tactic-shortname> and attack.t1059.001; also attack.g0016, attack.s0002. Checked 2026-11-14: sampled rules in SigmaHQ/sigma master already use attack.defense-impairment with attack.t1685 / attack.t1685.005, and the Atomic Red Team index uses stealth and defense-impairment. The Sigma tags appendix (v2.1.0, 2025-08-02) still lists defense-evasion, so check the target repo's validators before bulk retagging. attack sigma reports legacy tags, revoked IDs, and tactic mismatches per rule.
  • Rate robustness, not only presence: CTID Summiting the Pyramid (Sigma stp.* tags) scores how hard an analytic is to evade.
  • Coverage numbers overstate reality. Report rule count, tested count, and log source availability separately.

7. Threat intelligence

  • Profile: attack group <alias> gives techniques (direct and via attributed campaigns), software, and campaigns. Group pages lag current reporting; note the group object's modified date and add recent vendor reporting with citations.
  • Aliases are vendor-specific and overlap imperfectly. Say "tracked by MITRE as G0016 (APT29)" and keep vendor names as reported.
  • Overlap: attack layer --from A B C --op intersection for common tradecraft worth detecting first; union scores for sector heat maps. Shared techniques do not prove shared identity.
  • Prioritize by: groups targeting the organization's sector and region, then techniques seen in recent campaigns (C####), then prevalence (coverage gap ranking).
  • Software-heavy actors (and most Mobile/ICS groups) need --with-software to show real technique use.

8. Purple teaming and cross-framework links

ResourceUseNote
Atomic Red Team (redcanaryco/atomic-red-team)Small per-technique tests; attack atomic <ID>Enterprise only; already on v19+ IDs. Names and GUIDs only in outputs
MITRE Caldera (github.com/mitre/caldera)Automated adversary emulation with ability-to-technique mappingREADME now references apache/caldera; confirm current home
CTID Adversary Emulation Library (center-for-threat-informed-defense/adversary_emulation_library)Full and micro emulation plans per actorCheck which ATT&CK version each plan used
D3FEND (d3fend.mitre.org)Defensive countermeasures mapped to ATT&CK techniquesUse for "what control counters this"
MITRE Engage (engage.mitre.org)Adversary engagement, deception, and denial activitiesPair with emulation and detection plans
ATLAS (atlas.mitre.org)Attacks on AI/ML systemsPair with T1698-T1700, T1608.007, T1685.007
CAPEC / CWEAttack patterns and weaknessesSome techniques carry capec external references
CTID Mappings Explorer (ctid.mitre.org/mappings)NIST 800-53, cloud platform controls, CVE and KEV to ATT&CKEach mapping set is pinned to an ATT&CK version: check it, then validate the IDs
CISA KEV catalogExploited CVEs for prioritizing T1190 / T1203 / T1068-type gapsMap CVE -> technique via Mappings Explorer, not by guess
Mitigations (M####)attack get <technique> lists them; attack get M#### lists techniques coveredLink to NIST 800-53 through Mappings Explorer

Purple team loop: pick techniques from the gap list -> select Atomic tests -> run in authorized scope -> record which detection fired -> update detections.csv status to tested -> regenerate the layer.

9. Output templates

Incident / report mapping (header line: "ATT&CK Enterprise v20.1, mapped <date>, source <doc>")

#Time (UTC)Evidence (redacted quote + source)TacticTechnique IDTechnique nameConfidenceNotes

Follow with: unmapped behaviors and why, assumptions, validation result (attack validate: PASS).

Coverage report

  1. Scope: ATT&CK version, domain, platforms, rule sources, date.
  2. Summary: techniques with rule or better (n of N), tested (n), by tactic table (from coverage).
  3. Top gaps table: Technique | Tactic | Groups using | Required data components | Log source present? | Recommended action | Owner.
  4. Tag hygiene issues (revoked IDs, legacy tactics). 5. Navigator layer file. 6. Limits: coverage is not efficacy.

Emulation plan

  1. Objective, authorized scope, rules of engagement, ATT&CK version, actor (G####/C####) with sources.
  2. Step table: # | Tactic | Technique ID and name | Test reference (Atomic GUID or Caldera ability) | Platform | Expected telemetry (DC / log source) | Expected detection | Result.
  3. Safety: test accounts and hosts only, cleanup steps, stop conditions. 4. Findings feed back into detections.csv.

Sources

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.