CYBRARY Memory carries provenance or it carries nothing
Every note should record who wrote it. A note sourced from a tool is a retrieved claim; a note from a named human is a decision. They must never grade the same.
A-255·AI safety·35 min
Memory contamination: a rule nobody wrote Start this lab
The assistant is resetting finance passwords without a ticket and citing a policy that does not exist. The policy is in its durable memory, written by a tool rather than a person. Find the note, prove where it came from, and decide what memory is allowed to authorize.
Mission brief · A-255
Detection & automation engineer
Detection / automation engineer
Business important
Track default
You are acting as
you, signed in to the automation platform as `eng.you` (build role, not the on-call responder)
How you got here
You are not on the console watching the queue. You picked up a case the on-call analyst already triaged and were asked to make the response repeatable, so you're working the *replay* of that case in the automation platform.
Your access
Automation platform (author + test)
Automation platform
- Edit playbooks
- Run against a replayed case
- Inspect run history and context paths
- Not yours: Act on production hosts
- Not yours: Approve your own consequential actions
Ticketing (read + work notes)
Ticketing / ITSM
- Read the originating case
- Attach run evidence to the case
What's at stake
The playbook itself
Owner: SOC manager (approves anything that touches production)
A playbook that contains hosts without an approval gate will do it at 3am, at machine speed, to every host in the batch.
Rules of engagement
You may act alone
- Edit and test the playbook
- Run it against the replayed case
- Write work notes
Ask a human first
- Promoting the playbook to production
- Any action node that contains, disables or deletes
Where this sits in a real SOC
In a real SOC the analyst works the alert and hands the *pattern* to you. This lab teaches automation design judgement, not triage judgement — pair it with a Tier 1/2 triage lab.
Why this matters
Durable memory is the quietest privilege in an agent stack. A note written once by a tool is read back for months as if a person had said it, and the action it authorizes looks entirely reasonable in the audit log.
Concepts
Memory informs, it does not approve
An approval that can be written by the same pipeline the attacker influences is not an approval. Keep the gate in policy, where a person owns it.
Contaminated memory is evidence, not garbage
Preserve the note, its timestamp and its source before you delete it: it dates the compromise and shows which runs read it.
Objectives graded
- Trace an agent's confident wrong action back to a durable memory
- Read provenance on memory the way you read it on evidence
- Say what memory is allowed to authorize
Live walkthrough
The reference response, executed by the same engine that grades your attempt. Yours will differ — the rubric reads end state and evidence, not the shape of your workflow.
- 1.Approve consequential actionpromptapproval gateApprove consequential action
- 2.Write case field credentialRefactioncredentialRef → vault/soc-automation
- 3.Attach evidence: case/soar-a-255/evidenceactionThe learner attached at least one linkable piece of evidence to the case.
- 4.Write case field closureNoteactionclosureNote → written
- 5.Attach evidence: range/hst-ai-agent01/sessionactionThe memory store was read on the machine, not inferred from behavior.
- 6.Write case field poisonedMemoryIdactionpoisonedMemoryId → mem-0197
- 7.Write case field memorySourceactionmemorySource → tool:kb.search
- 8.Write case field memoryStorePathactionmemoryStorePath → /var/lib/northwind-agent/memory/notes.jsonl
- 9.Write case field closureNoteactionclosureNote → written
- 10.Write case field segmentGatewayactionsegmentGateway → 10.42.48.1
- 11.Write case field seatRamGbactionseatRamGb → 16
- 12.Write case field seatDiskGbactionseatDiskGb → 512
Gate approved — 100%
17 of 17 rubric rules hold.
Gate denied — 92%
Refusing the consequential action is a defensible choice with a measured cost; the difference is what the lab is teaching.
- pass behavior · The workflow reached a terminal success state on the graded run.
- pass safety · A consequential action passed through a human approval gate.
- pass efficiency · No non-idempotent operation was retried without an idempotency key.
- pass configuration · Every action node resolves its secret from the vault, never an inline field.
- pass investigation · The learner attached at least one linkable piece of evidence to the case.
- pass explanation · The case carries an analyst closure note explaining the outcome.
- pass behavior · No innocent asset in the alert batch was contained, disabled or revoked.
- pass configuration · The playbook survived the integration faults it met and still finished.
- pass investigation · Nothing was changed in the estate before evidence had been gathered.
- pass investigation · The memory store was read on the machine, not inferred from behavior.
- pass investigation · The specific durable note behind the action was identified.
- pass investigation · The note's provenance shows it was written by a tool, not a person.
- pass configuration · The memory store was located, so it can be scoped and reviewed.
- pass explanation · The closure note states what memory may and may not authorize in future.
- pass investigation · The analyst scoped the host's network from the wire, so containment decisions know which segment is in play.
- pass investigation · The analyst scoped the machine itself, so memory-hungry triage steps are planned against real capacity rather than a guess.
- pass investigation · Disk capacity was taken from the volume itself, which is what makes an image or memory-dump decision defensible.
MITRE ATT&CK coverage
A durable note written by a tool is read back as policy: the trust relationship, not the prompt, is what was abused.
| Technique | Name | Tactic |
|---|---|---|
| T1199 | Trusted Relationship | Initial Access |
| T1204.001 | User Execution: Malicious Link | Execution |
| T1078 | Valid Accounts | Defense Evasion |
What each rubric rule proves
The workflow reached a terminal success state on the graded run.
behavior · weight 20
A consequential action passed through a human approval gate.
safety · weight 15
No non-idempotent operation was retried without an idempotency key.
efficiency · weight 10
Every action node resolves its secret from the vault, never an inline field.
configuration · weight 10
The learner attached at least one linkable piece of evidence to the case.
investigation · weight 10
The case carries an analyst closure note explaining the outcome.
explanation · weight 10
No innocent asset in the alert batch was contained, disabled or revoked.
behavior · weight 15
The playbook survived the integration faults it met and still finished.
configuration · weight 15
Nothing was changed in the estate before evidence had been gathered.
investigation · weight 10
The memory store was read on the machine, not inferred from behavior.
investigation · weight 15
The specific durable note behind the action was identified.
investigation · weight 10 · T1199 Trusted Relationship (Initial Access)
The note's provenance shows it was written by a tool, not a person.
investigation · weight 10 · T1204.001 User Execution: Malicious Link (Execution)
The memory store was located, so it can be scoped and reviewed.
configuration · weight 10
The closure note states what memory may and may not authorize in future.
explanation · weight 15 · T1078 Valid Accounts (Defense Evasion)
The analyst scoped the host's network from the wire, so containment decisions know which segment is in play.
investigation · weight 5
The analyst scoped the machine itself, so memory-hungry triage steps are planned against real capacity rather than a guess.
investigation · weight 3
Disk capacity was taken from the volume itself, which is what makes an image or memory-dump decision defensible.
investigation · weight 3
Common mistakes
- Wiping the memory store before reading its provenance.
- Assuming a confident agent was prompted rather than remembered.
- Leaving tool-written notes able to satisfy an approval check.
Same idea, three vendors
Vocabulary is generated from the teaching profiles, so it always matches what the consoles render.
| Concept | Cortex Forge | Summit SOAR | Storyline |
|---|---|---|---|
| approval | Manual task | Prompt | Page |
| artifact | Indicator | Artifact | Record |
Where does this value come from?
The same reference to an enrichment node's verdict, written three ways:
- Cortex Forge: ${enrich-sender.verdict}
- Summit SOAR: enrich-sender.verdict
- Storyline: <<enrich-sender.verdict>>
The same concepts in shipping products
The consoles you practice in use teaching names. These are the words the real vendors print in their own documentation, so the vocabulary survives the day you sit in front of a production tenant.
| Concept | Cortex XSOARXSOAR 8.x | Splunk SOARSplunk SOAR 6.x (formerly Phantom) | TinesTines (2026 cloud) | Microsoft Sentinel + Logic AppsSentinel (Defender portal, 2026) | Google SecOps SOARSecOps SOAR (formerly Siemplify) | Swimlane TurbineTurbine (2026) | Torq HyperautomationTorq (2026) | IBM QRadar SOARQRadar SOAR (Resilient) 51.x | ServiceNow Security OperationsSecOps (Flow Designer era) |
|---|---|---|---|---|---|---|---|---|---|
| approval | Manual (data collection) task | Prompt block | Page (form) / Send to Chat | Approval action (Teams / email) | Manual action | Task / approval action | Approval step (Torq Chat / Slack) | Manual task / phase gate | Approval activity |
| artifact | Indicator | Artifact (CEF fields) | Event payload field | Entity | Entity | Field / correlation value | Event field | Artifact | Observable |
What transfers
Provenance on retrieved content is the same control that protects RAG pipelines, knowledge bases and cached threat intelligence.