SaaS ✓ fields verified against sample logs

GitHub Audit Log Fields: action, actor and More, With Examples

Published

One JSON event per action in a GitHub organization: the action as category.operation (repo.create, org.add_member, hook.create…), who did it (actor), in which organization and repository, when (epoch milliseconds), and from which address when IP disclosure is enabled.

Official documentation: GitHub: Audit log events for your organization · GitHub: Reviewing the audit log for your organization

Sample record

[{"action":"repo.create","actor":"alice","org":"corp-org","repo":"corp-org/new-feature","created_at":1705309200000,"actor_ip":"203.0.113.10","@timestamp":1705309200000}

From the LogTriage parser test corpus — synthetic data, no real hosts or users.

Explain this record field by field →

Field reference

Field Type Meaning Why it matters for security
actione.g. repo.destroy string GitHub: "The name for each audit log entry is composed of a category of events, followed by an operation type. For example, the repo.create entry refers to the create operation on the repo category." repo.destroy, org.add_member, hook.create, org.disable_two_factor_requirement and protected_branch.destroy are the actions an attacker with owner access uses.
actore.g. alice string The user account that performed the action. A token or account acting outside its usual hours or from a new address is the first thing to check after a leaked personal access token.
orge.g. corp-org string The organization the event belongs to. Filter on it when one export covers several organizations.
repoe.g. corp-org/api-service string The repository the action affected, as owner/name. Deletions and visibility changes on important repositories are the high-impact ones.
@timestampe.g. 1705328200000 number (epoch ms) When the event happened, in milliseconds since the Unix epoch. Divide by 1000 before converting; tools that expect seconds put events in the year 56,000.
created_ate.g. 1705309200000 number (epoch ms) GitHub: "Timestamp shows the time since Epoch with milliseconds." The same instant as @timestamp in exported events. Search the audit log by it with the created qualifier; older events need an explicit date range.
actor_ipe.g. 185.220.101.42 IP address The address the actor came from, present only when the organization or enterprise has enabled display of IP addresses in the audit log. The same address behind a 2FA change, a new member and a new webhook is a takeover; without IP disclosure enabled you cannot see it.
usere.g. new-backdoor-account string The user the action was performed on — the member added, removed or changed — as distinct from the actor. An actor adding an unknown user as a member right after a security change is persistence.
hook_ide.g. 123456 number The webhook the hook.* event refers to. A new webhook can send every push, including code and secrets, to an outside server; check its URL in the organization settings.

Have one of your own records? Paste it into the log line explainer and each field is explained from this table, with times, codes and addresses decoded — in your browser.

What it records

The organization audit log records what members, apps and tokens did to the organization: repositories created, deleted or made public, members added, teams changed, webhooks and deploy keys created, security settings changed. Each event’s action is a category and an operation (repo.destroy), and GitHub lists every action and the fields it carries in Audit log events for your organization. Common fields are the actor, the org, the repo, the user acted on, and the timestamps @timestamp and created_at in epoch milliseconds; actor_ip appears when IP disclosure is enabled.

The malicious sample is an organization takeover from two addresses in three minutes: a repository deleted, two-factor authentication disabled, a new member added and a webhook created — the webhook being the attacker’s way to keep receiving code after access is cut.

Where to get it

  • Organization settings → Audit log. Search by action, actor or repository and export as JSON or CSV. Only owners can see it.
  • REST API. The audit log endpoint returns the same events for scripted collection (GitHub Enterprise Cloud).
  • Audit log streaming. Enterprise accounts can stream events continuously to cloud storage or a SIEM.
  • Git events. Clone, fetch and push events are included only in some exports and API queries; check what your export contains.
# Actions per actor and address, most frequent first, from a JSON export
jq -r '.[] | [.actor, (.actor_ip // "-"), .action] | @tsv' export.json | sort | uniq -c | sort -rn | head -20

Parsing pitfalls

  • Milliseconds. @timestamp and created_at are epoch milliseconds.
  • Fields vary by action. GitHub lists a different field set per action; hook_id exists on hook events, team on team events.
  • Actor vs user. Read both; the target of an action is in user, team or repo.
  • IP addresses are optional. actor_ip is missing unless IP disclosure was enabled when the event happened.
  • 180 days. Older events are gone unless you streamed or exported them.

Detection ideas

  • Security controls removed: org.disable_two_factor_requirement, protected_branch.destroy, secret scanning or Dependabot disabled — the GitHub org takeover rule.
  • Persistence: org.add_member, hook.create, new deploy keys or OAuth app approvals right after a security change.
  • Destruction and exposure: repo.destroy, repositories made public, or transfers out of the organization.
  • Token misuse: many actions from one actor at an unusual time or from a new actor_ip.

The free browser analyzer reads GitHub audit log exports and runs these checks locally.

Sending GitHub Audit Log to an LLM

One megabyte of this format is about 490,584 tokens on current Claude models and 340,898 on GPT models, measured with the models' own tokenizers — so a gigabyte sent raw to Claude Sonnet 5.5 costs about $981 in input tokens alone (prices of 2026-10-24). Estimate your own volume →

Detections and guides for GitHub Audit Log

Frequently Asked Questions

How long does GitHub keep the organization audit log?
GitHub states that the audit log contains events from the last 180 days, and the web view shows the last three months unless you search with a date range. To keep events longer, stream the audit log (available on GitHub Enterprise) or export it regularly through the REST API.
Why is there no IP address in my GitHub audit log?
Source IP addresses appear only when display of IP addresses in the audit log is enabled for the organization or enterprise; GitHub logs enabling and disabling it as org.enable_source_ip_disclosure and org.disable_source_ip_disclosure. Turn it on before an incident, because past events do not gain addresses afterwards.
What is the difference between actor and user?
actor is who did it; user is who it was done to. In org.add_member the actor is the owner who added someone and user is the new member. Rules that only read actor miss accounts an attacker adds for later use.
How are GitHub audit log timestamps written?
As milliseconds since the Unix epoch, in @timestamp and created_at; GitHub's documentation describes created_at that way. Convert by dividing by 1000. The audit log search uses the created qualifier with dates, not these numbers.

Analyze a GitHub Audit Log file in your browser

The free analyzer detects this format, parses every field above and flags attack patterns — the file never leaves your machine. For threat-intel enrichment and an AI-written incident report, run it through LogTriage.

Investigating something right now? LogTriage checks every address in the file against threat intelligence and writes the incident report: 10 analyses a month free. The incident pack adds the in-depth AI analysis and files up to 100 MB: 25 analyses for $19, one payment, no subscription. Pricing

How this page was made. The official definitions, field names and examples on it are copied by script from the sources it links and re-checked against them on every build. The investigation advice was drafted with AI from those sources and passed a separate review before publishing. What is checked, and how · found a mistake? [email protected]