// verdict · checked 2026-08-11 · Copilot cloud agent (formerly 'coding agent'), 2026 docs
Copilot coding agent
GitHub
Assigned an issue like an intern, ships a draft PR like an intern.
- CAPABILITY
- 4/5 AGENT
- PROOF
- OPERATOR INSPECTABLEYou can run it and read the steps yourself — the record exists for whoever runs it
- TRACE ACCESS
- OPERATOR-OWNEDWhoever runs it holds the record; nothing is published
- LEASH
- Not measuredno execution record scored — measure it yourself
| LOOPSruns until doneyes | Runs multi-step in its own ephemeral Actions VM — explores repo, plans, edits, runs tests, pushes commits iteratively; step count varies by task, with only a 1-hour session timeout as ceiling. |
| CHOOSESpicks its own toolsyes | Docs state it "will use available tools autonomously, and will not ask for approval before use"; it decides which commands, tests and linters to run as it explores the repo. |
| ACTSchanges the worldyes | It automates branch creation, commit message writing and pushing, and opens a draft pull request — real repository state change, not text for a human to apply. |
| RECOVERSfixes its own errorsyes | Its own test/lint runs plus CodeQL, secret scanning and code review feed back in and it "attempts to resolve issues identified prior to completing the pull request" with no human. |
| UNSUPERVISEDnobody watchingno | Docs: workflows wait for a human to click 'Approve and run workflows'; it can't approve or merge its PR. |
Capability decisions combine documented behaviour and execution evidence. The proof label on each row shows which of the two decided it, and Leash is reported separately, only when it was measured from an execution record. Primary source: docs.github.com. Tested against: Copilot cloud agent (formerly 'coding agent'), 2026 docs. Our confidence in this dossier: 6/10.
Can you see it work?
the weakest of the five grades aboveOPERATOR INSPECTABLE — You can run it and read the steps yourself — the record exists for whoever runs it. Trace access: OPERATOR-OWNED — Whoever runs it holds the record; nothing is published.
PR links to session logs showing Copilot's reasoning and the tools it used
Where we looked: GitHub Docs 'manage and track agents' page and the github/docs source tree (content/copilot/concepts/agents/cloud-agent/about-cloud-agent.md) via the GitHub API. · see the evidence →
No revisions yet. This verdict has stood since 2026-08-11.
Badges
earned, never for sale
These render from the live dataset, so a badge cannot outlive the verdict it claims. If the verdict changes the badge changes with it, and an unearned one returns 409 rather than an image. No sponsor can buy one, at any price.
Think this is wrong?
bring evidence, not opinionName the criterion and link a trace, a doc, or a recorded run that shows the behaviour. A verdict changed by evidence is the best thing that can happen to this site, and the change goes on the public record with your reason attached. Vendors are welcome, and a vendor's own filing is labelled as such rather than buried.