# My agent trace review

LockedIn Labs · LockedIn FDE Training · Editable practice sheet · Edition 1

Use this sheet with the free [25-minute trace-design exercise](https://training.lockedinlabs.ai/training/guides/workbook#optional-lab-design-an-agent-trace-without-copying-the-payload-25-minutes).
Read its eight synthetic proposals before writing. This file contains no completed
review, collected telemetry or approval. Work locally with the supplied fictional
case; include no real requests, credentials, private source text or client records.
Reading or downloading it does not save work to the platform.

## 01 — State the investigation

- Practice revision/date: [your local revision and date]
- Mode: [tabletop; identify a separate executed experiment only if one occurred]
- Receiver: [practice role; peer or self-review]
- Question: [what must the receiver decide?]
- Packet used: [workbook section and P1–P8; supplied observations, not your own run]
- Reader boundary: [which role may inspect which scoped metadata?]
- Content excluded from this store: [fields and reason; no sample secrets]
- Retention decision still needed: [responsible role and unresolved requirement]

Use the exercise's suggested 5/7/8/5-minute phases after reading the packet.
Pause for unfamiliar terms and record extra time rather than treating the estimate
as a deadline or a completed learning outcome.

## 02 — Review all eight proposals

Choose retain, transform or exclude for the **support event store**, then explain
the choice. Do not copy raw content into the table. Preserve the useful security
event even when an unsafe field is excluded. Blank means unresolved, not passed.

| Packet ID | My first decision | Trusted origin or provenance gap | Safe diagnostic to retain | Evidence still needed |
| --- | --- | --- | --- | --- |
| P1 | [retain / transform / exclude] | [origin] | [bounded fields] | [gap] |
| P2 | [decision] | [origin] | [bounded fields] | [gap] |
| P3 | [decision] | [origin or mismatch] | [bounded fields] | [gap] |
| P4 | [decision] | [origin] | [bounded fields] | [gap] |
| P5 | [decision] | [origin] | [bounded fields] | [gap] |
| P6 | [decision] | [origin] | [bounded fields] | [gap] |
| P7 | [decision] | [origin] | [bounded fields] | [gap] |
| P8 | [decision] | [origin] | [bounded fields] | [gap] |

Before opening a worked interpretation, identify one row where plausible prose
and an application observation disagree. Explain which claim the packet supports
and which observation would resolve the mismatch.

My first explanation: [write here before consulting help]

## 03 — Repair P1 and P4 without inventing observations

Write two structured proposals. Use supplied values where available. Use null
for a missing observation and name it in the gap register; never invent a time,
revision, scope or adapter mode. Bracketed strings are editable placeholders.
This is a draft contract, not deployable code. A null does not establish that a
production schema accepts it. Omit fields not permitted by your proposed contract.

```json
[
  {
    "requestId": "[P1 request]",
    "attemptId": "[P1 attempt]",
    "scopeRef": null,
    "eventTime": null,
    "applicationRevision": null,
    "stage": "[application-observed stage]",
    "outcome": "[application-observed outcome]",
    "reason": "[proposed bounded reason]",
    "durationMs": null,
    "adapterMode": null
  },
  {
    "requestId": "[P4 request]",
    "attemptId": "[P4 attempt]",
    "scopeRef": null,
    "eventTime": null,
    "applicationRevision": null,
    "stage": "[application-observed stage]",
    "outcome": "[application-observed outcome]",
    "reason": "[proposed bounded reason]",
    "durationMs": null,
    "adapterMode": null
  }
]
```

| Record / field | Supplied fact or missing observation | Authorized source to inspect | Responsible role / next action |
| --- | --- | --- | --- |
| [P1 or P4 / field] | [packet reference or unknown] | [bounded lookup, not a raw payload dump] | [role and action] |

Define who owns each proposed reason label. Separate confirmed outcomes from
suspected causes. State which interval each duration measures; an unattempted
stage has no measured latency.

## 04 — Give the receiver a useful next step

An absent log does not prove that nothing happened. Keep attempted emission,
a durable queue receipt and collector acknowledgement distinct.

| Condition | What is established | What remains unknown | Next permitted investigation / owner |
| --- | --- | --- | --- |
| P1 denial | [bounded supplied observation] | [remaining question] | [action / role] |
| P4 timeout | [bounded supplied observation] | [cause and useful task status] | [action / role] |
| P7 collection uncertainty | [bounded supplied observation] | [delivery state] | [action / role] |

- Failure signal without payload leakage: [bounded fields and destination]
- Buffer/retry limit to specify: [limit and owner; not a claim it exists]
- Required-audit failure policy still needed: [decision role and evidence]
- P6 serialization/viewing risk: [how untrusted text stays data]
- P8 mode statement: [what was supplied, and what was not observed]
- P5 changed condition: [receiver access is revoked; historical reference vs new lookup]
- Useful work permitted to continue: [agreed policy and scope, or unresolved]

## 05 — Preserve the receiver's challenge and your revision

Give the receiver your proposed records and table without excluded content.
Ask for the next investigation for denial, timeout and unconfirmed collection,
and the revoked-access decision. Do not answer for them. Label solo work self-review.

| Receiving task | Actual action / question | Assistance supplied | Result or unobserved | Revision needed |
| --- | --- | --- | --- | --- |
| [one task] | [what happened] | [specific help or none] | [actual limit] | [smallest repair] |

- Original claim I changed: [retain its exact earlier wording]
- Evidence or feedback that changed it: [packet ID / peer observation / authored reference]
- Revised claim and reason: [bounded conclusion]
- Action repeated after revision: [actual observation, or not attempted]
- Remaining gap, owner and next evidence: [specific unresolved item]

Retain the first attempt and revision beside portfolio section 05. With existing
Academy access, open the [related handoff lesson](https://training.lockedinlabs.ai/app/academy?course=fde-client-delivery&lesson=deployment-and-handoff&view=focus)
for worked interpretation and a targeted check. A local file is not a saved
playbook entry, reviewed submission, implemented logger, qualification or client
acceptance. Use the normal authorized review route if you choose to submit work.