Skip to content

feat(debug): add persistent website diagnostics and network controls - #295

Open
iuyo5678 wants to merge 9 commits into
mainfrom
feat/agent-website-debugging
Open

iuyo5678 wants to merge 9 commits into
mainfrom
feat/agent-website-debugging

Conversation

@iuyo5678

Copy link
Copy Markdown
Collaborator

Summary

Developers investigating their own websites need a retained evidence trail that an agent can inspect without stitching together unrelated console, network, and page snapshots. This adds an opt-in debugging workflow across the CLI, DSH tools, and extension, while preserving the existing popup controls.

  • Capture requests and bounded text bodies, console events, page context, and manual or agent operations. Group evidence around each operation, expose field changes, and identify missing or partial capture explicitly.
  • Save browser-local history with detail views and JSON export. Add bounded query output, filters, capability discovery, request pinning, and explicit busy-state/wait commands.
  • Record native main-frame performance evidence with validity reasons and visibility history. Add endpoint aggregation, slow-request summaries, and suspected duplicate-request statistics with source request IDs.
  • Add task-scoped request blocking, modification, mock responses, and same-origin replay. Retain intervention provenance and clean up controls when capture or ownership ends.
  • Harden capture deadlines and cancellation, retain in-flight requests beyond the recent-request cache, preserve legacy history details, and allow the original task to read saved captures after they leave the four-run live cache.

Scope and limits

Capture starts only when requested and is not retroactive. Persistent evidence and response bodies remain bounded; redaction, truncation, interruption, and storage gaps are reported. Task and tab ownership restrict access. Performance support is main-frame only and excludes INP and CPU profiling. Temporal associations and suspected duplicates are investigation aids, not causal conclusions. Source-code repair and automatic repair verification are outside this feature.

The extension adds current-task access and a dedicated debugging/history workspace. Existing popup functions remain available. Documentation, schemas, agent instructions, and translations are included.

Validation

  • Extension suite: 1,843 passed; 102 environment-dependent tests skipped.
  • Isolated Chrome debugging suite: 96 passed, including real network controls, performance capture, IndexedDB recovery/upgrade, frozen-renderer deadlines, and late-hook recovery.
  • DSH suite: 245 passed. Rust workspace: 715 passed on the latest CLI/protocol changes; the final hardening commit changes extension code only.
  • TypeScript, lint, Rust formatting/Clippy, production builds, and extension package consistency checks passed.
  • Regression coverage includes slow requests surviving cache pressure, pending body reads, explicit tracking overflow, legacy body/header/timing detail reads, more than four captures in one task, reused task IDs, and revoked tab access.

Related: #222 and #61.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants