Skip to content

thiefdroppers/cyber-investigator-skills / ai-agent-skills/industries/msp-incident-response

cyber-investigator-msp-incident-response

Verified

Applies the Cyber Investigator Roadmap method to incident response that managed service providers (MSPs) and small incident-response consultancies run for small and mid-sized business (SMB) clients. Use when asked to triage a suspected business email compromise or mailbox takeover in Microsoft 365 or Google Workspace, review a client tenant's audit logs for signs of compromise, lay out what data was accessed and when, or draft an insurer-facing incident summary. Authorized, defensive, lawful scope only.

FileKindSize
SKILL.mdskill md7.6 KB
reference/email-compromise-triage.mdmarkdown4.4 KB

reference/email-compromise-triage.md

View on GitHub
# Email-compromise triage (Microsoft 365 and Google Workspace)

This is the triage procedure that [`../SKILL.md`](../SKILL.md) points to for a suspected business email compromise or mailbox takeover. Work through the steps in order. Each step checks one account at a time, so repeat steps 2 through 6 for every account in scope.

Before step 1, confirm the engagement authorization and the list of accounts in scope, as the skill's scope section requires. Investigate through the admin tools and logs only. Never sign in to the compromised account as the user.

For every data point, record the value, the log or admin screen it came from, and the date and time you pulled it. Record the time zone of each log, because admin consoles and exported logs do not always use the same one.

## Procedure: email-compromise triage
1. Preserve first. Before anyone resets a password, deletes a rule, or removes an app, export the audit log and the message trace for the accounts in scope and the period in question. In Microsoft 365 that is the audit log search and the message trace. In Google Workspace it is the audit and investigation tool and the email log search. Save the exports unchanged, and record when and by whom each was pulled. Log retention is limited and depends on the client's license, so check the retention window for this tenant and note in the case file if it does not reach back to the suspected start of the compromise.
2. Check inbox rules. List every inbox rule and mail-forwarding setting on the account. Flag any rule that forwards or redirects mail to an outside address, deletes messages, marks them read, or moves them to a folder the user rarely opens, such as RSS or archive folders. Rules that match on words like "invoice," "payment," "wire," or a specific contact's name are a common way to hide the attacker's conversation from the user. For each flagged rule, record its conditions, its actions, and when it was created, and find the creation event in the audit log.
3. Check OAuth app consents. List the third-party apps the account has granted access to, with the permissions each holds and when the consent was granted. Flag any app the user does not recognize, any app with mail read or send permissions granted during the suspected window, and any app granted access by one user that appears nowhere else in the tenant. A consented app keeps its access after a password reset, so it has to be found and recorded before containment.
4. Check MFA state and recent changes to it. Record whether MFA is enforced on the account, which methods are registered, and when each was added. Flag any method added, removed, or changed during the suspected window, especially a new phone number or authenticator device. Check whether the tenant allows legacy authentication protocols that skip MFA, and record the answer.
5. Check sign-in geography and timing. Pull the sign-in log for the account and compare each sign-in against the user's normal locations, devices, and working hours. Flag sign-ins from an unexpected country or network, two sign-ins from distant locations too close together for travel, and sign-ins from a new device or client just before a rule or app consent appears. A VPN or a mobile carrier can make a legitimate sign-in look foreign, so confirm with the client before treating a location alone as a finding.
6. Check mailbox-delegation changes. List who has delegate, full-access, send-as, or send-on-behalf permissions on the mailbox, and who the account has delegate access to. Flag any permission added during the suspected window, and find the matching admin or audit event showing who added it.
7. Build the timeline. Put every flagged event from steps 2 through 6 in time order with its source. The first confirmed attacker action marks the start of the known compromise window. If something in the logs suggests an earlier start that cannot be confirmed, record it as unconfirmed rather than moving the start.
8. State what was accessed. From the audit log, list the mailbox items, files, or records that the logs show were read, sent, downloaded, or changed during the window. If the tenant's logging does not record item-level access, write that the data accessed is not determined and say why. Do not estimate.
9. Report the findings with the report format in [`../SKILL.md`](../SKILL.md). Tag each finding confirmed (two or more log sources agree) or likely (one source only), as the generic recon cycle does, and never report a likely finding as confirmed.
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: 472kUpdated Claude CodeCodexVerifiedInstall
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: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 196kUpdated Claude CodeCodexVerifiedInstall