Browser: add ChatGPT conversation import

[mustafa0x]

Summary

• add oracle import-chatgpt-url for registering manual ChatGPT conversation URLs as completed browser sessions
• store sanitized browser metadata so normal --followup can resume through the existing fail-closed path
• document the flow and add unit/integration coverage

Verification

• pnpm exec vitest run tests/cli/importChatgptConversation.test.ts tests/cli/followup.test.ts tests/cli/integrationCli.test.ts
• pnpm run check
• pnpm run docs:check
• pnpm run build
• autoreview clean: no accepted/actionable findings reported

[clawsweeper]

Codex review: needs real behavior proof before merge. Reviewed August 1, 2026, 1:03 PM ET / 17:03 UTC.

ClawSweeper review

What this changes

This PR adds oracle import-chatgpt-url to register a validated manually started ChatGPT conversation as a browser session that can later continue through --followup.

Merge readiness

⛔ Blocked until real behavior proof is added - 4 items remain

Keep this PR open for maintainer review rather than cleanup: current main has the fail-closed browser follow-up boundary but no equivalent manual conversation-import command, and the proposed implementation is focused with automated coverage. It still lacks the required after-fix proof of importing a real signed-in ChatGPT thread and successfully submitting its first follow-up; the new public workflow also needs maintainer acceptance.

Priority: P3
Reviewed head: fb7bb7c07f01b9f3fc2cc3b022507347cb0403e3
Owner decision: Required. See Decision needed.

Review scores

Measure	Result	What it means
Overall readiness	🧂 unranked krab (1/6)	The implementation is focused and covered by targeted tests, but it is not merge-ready without real browser proof for this persisted-session workflow.
Proof confidence	🧂 unranked krab (1/6)	Needs real behavior proof before merge: The PR body reports automated checks only; add redacted after-fix terminal output, logs, or a recording that shows a real signed-in conversation import and its first successful --followup, without exposing conversation contents, URLs, IP addresses, cookies, or credentials. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Verification

Check	Result	Evidence
Real behavior	Needs proof	Needs real behavior proof before merge: The PR body reports automated checks only; add redacted after-fix terminal output, logs, or a recording that shows a real signed-in conversation import and its first successful --followup, without exposing conversation contents, URLs, IP addresses, cookies, or credentials. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.
Evidence reviewed	5 items	Current resume boundary: Current main resolves a browser follow-up only through a recoverable ChatGPT conversation URL and resets the child run to non-research, non-archiving behavior; this is the safety boundary the PR reuses. No current-main equivalent: A repository-wide search of CLI, source, docs, README, and tests found existing follow-up/recovery support but no import-chatgpt-url command or manual-conversation import path. Proposed session semantics: The branch creates a browser session marked completed, stores the supplied conversation URL and ID as runtime/harvest metadata, and relies on the normal follow-up gate for the first actual browser interaction.
Findings	None	None.
Security	None	None.

How this fits together

Oracle browser sessions persist the ChatGPT conversation identity and browser configuration after a run. A later --followup resolves that stored identity, reopens the conversation in a signed-in browser, and refuses to submit if the URL or prior turns are not safe to resume.

flowchart LR
  A[Manual ChatGPT conversation URL] --> B[Import command]
  B --> C[Stored browser session metadata]
  C --> D[Follow-up command]
  D --> E[Signed-in URL and prior-turn checks]
  E --> F[Continue conversation or fail closed]

Loading

Decision needed

Question	Recommendation
Should Oracle support manually importing a ChatGPT conversation URL as a first-class public browser-session workflow, with the imported session initially trusted only enough to reach the existing fail-closed follow-up checks?	Sponsor after live proof: Keep the narrow implementation, require a current-main rebase plus redacted import-and-follow-up evidence, then merge if the browser checks behave as documented.

Why: The patch adds a durable CLI workflow rather than repairing an established broken contract; maintainers need to choose whether this is a supported product surface after reviewing real browser behavior.

Before merge

• Add real behavior proof - Needs real behavior proof before merge: The PR body reports automated checks only; add redacted after-fix terminal output, logs, or a recording that shows a real signed-in conversation import and its first successful --followup, without exposing conversation contents, URLs, IP addresses, cookies, or credentials. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.
• Resolve merge risk (P1) - The command records an unverified external conversation as a completed local session; although the first follow-up remains fail-closed, maintainers need real signed-in browser proof that the actual thread is reopened and prior-turn validation works.
• Resolve merge risk (P1) - The branch predates current-main browser session lifecycle changes, so its automated results do not demonstrate compatibility with the current integration surface until it is rebased and rerun.

Agent review details

Security

None.

Review metrics

Metric	Value	Why it matters
Patch surface	10 files affected; 506 additions, 0 deletions	The feature crosses CLI parsing, persisted session metadata, documentation, and tests, so a current-main end-to-end proof matters more than unit coverage alone.

Merge-risk options

Maintainer options:

1. Rebase and prove the live resume (recommended)
   Refresh the branch against current main, then add redacted terminal output, logs, or a recording showing a real signed-in conversation import followed by a successful first --followup.
2. Decline the new session type
   Pause or close the PR if maintainers prefer browser follow-ups to remain limited to sessions created by Oracle itself.

Technical review

Best possible solution:

Keep the existing fail-closed resume contract, rebase this focused command onto current main, and accept it only after a redacted live import followed by a successful first --followup proves the end-to-end workflow.

Do we have a high-confidence way to reproduce the issue?

No. The branch tests metadata construction and resolution, but the PR provides no real signed-in browser run showing that a manually imported conversation passes the first follow-up's prior-turn checks and submits successfully.

Is this the best way to solve the issue?

Unclear. Reusing the existing recoverable-URL and prior-turn gate is the narrow maintainable design, but the command should not be accepted until maintainers confirm the new public workflow and live proof verifies it on current main.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 68b8c51b0ee0.

Labels

Label justifications:

• P3: This is an optional CLI/browser workflow enhancement rather than a regression in an existing supported path.
• merge-risk: 🚨 session-state: The PR adds a new persisted completed-session form whose metadata controls later browser conversation resumption.
• rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🐚 platinum hermit.
• status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR body reports automated checks only; add redacted after-fix terminal output, logs, or a recording that shows a real signed-in conversation import and its first successful --followup, without exposing conversation contents, URLs, IP addresses, cookies, or credentials. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence

What I checked:

• Current resume boundary: Current main resolves a browser follow-up only through a recoverable ChatGPT conversation URL and resets the child run to non-research, non-archiving behavior; this is the safety boundary the PR reuses. (src/cli/followup.ts:25, 68b8c51b0ee0)
• No current-main equivalent: A repository-wide search of CLI, source, docs, README, and tests found existing follow-up/recovery support but no import-chatgpt-url command or manual-conversation import path. (bin/oracle-cli.ts, 68b8c51b0ee0)
• Proposed session semantics: The branch creates a browser session marked completed, stores the supplied conversation URL and ID as runtime/harvest metadata, and relies on the normal follow-up gate for the first actual browser interaction. (src/sessionManager.ts:636, fb7bb7c07f01)
• Repository browser-proof policy: The repository manual-test guidance requires real browser results and session IDs in PR descriptions for browser automation changes; the PR body lists automated checks but has no redacted import-plus-follow-up runtime evidence. (docs/manual-tests.md:140, 68b8c51b0ee0)
• Relevant current-main provenance: Blame attributes the current recoverable-URL guard to the current-main baseline, while session lifecycle code was updated after this PR branch by a browser-capture fix; the branch should be refreshed before relying on its older integration evidence. (src/browser/reattachability.ts:10, 3f9db037ef6d)

Likely related people:

• Peter Steinberger: Current-main blame for the recoverable ChatGPT URL gate and follow-up path traces to the current baseline authored by Peter Steinberger. (role: current resume-path contributor; confidence: medium; commits: 5daa6ce8c352; files: src/cli/followup.ts, src/browser/reattachability.ts)
• Rokurolize: Rokurolize authored the post-branch browser-capture interruption fix that touches the session manager this PR extends. (role: recent session lifecycle contributor; confidence: medium; commits: 3f9db037ef6d; files: src/sessionManager.ts)

Rank-up moves

Optional improvements that raise the rating; they are not merge blockers.

• Rebase onto current main and rerun the focused CLI tests.
• Add redacted real browser evidence for import plus the first successful follow-up, then update the PR body to trigger re-review.

Rating scale

Score	Internal tier	Crab rank	Meaning
6/6	S	🦀 challenger crab	Exceptional readiness
5/6	A	🦞 diamond lobster	Very strong readiness
4/6	B	🐚 platinum hermit	Good normal PR; ordinary maintainer review
3/6	C	🦐 gold shrimp	Useful, but confidence is limited
2/6	D	🦪 silver shellfish	Proof or implementation needs work
1/6	F	🧂 unranked krab	Not merge-ready
N/A	NA	🌊 off-meta tidepool	Rating does not apply

Overall follows the weaker of proof and patch quality.
Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

Workflow

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

History

Review history (16 earlier review cycles; latest 8 shown)

• reviewed 2026-07-31T05:23:39.759Z sha fb7bb7c :: needs real behavior proof before merge. :: none
• reviewed 2026-07-31T11:27:41.124Z sha fb7bb7c :: needs real behavior proof before merge. :: none
• reviewed 2026-07-31T14:27:55.502Z sha fb7bb7c :: needs real behavior proof before merge. :: [P3] Remove the release-owned changelog entry
• reviewed 2026-07-31T15:34:40.368Z sha fb7bb7c :: needs real behavior proof before merge. :: [P3] Remove the release-owned changelog entry
• reviewed 2026-07-31T20:17:33.718Z sha fb7bb7c :: needs real behavior proof before merge. :: none
• reviewed 2026-07-31T21:40:56.871Z sha fb7bb7c :: needs real behavior proof before merge. :: none
• reviewed 2026-08-01T13:53:46.095Z sha fb7bb7c :: needs real behavior proof before merge. :: none
• reviewed 2026-08-01T15:49:05.824Z sha fb7bb7c :: needs real behavior proof before merge. :: none