diff --git a/authors/gaetano.md b/authors/gaetano.md
new file mode 100644
index 00000000..2e0bf202
--- /dev/null
+++ b/authors/gaetano.md
@@ -0,0 +1,13 @@
+Author: Gaetano
+Title: Software Engineer
+Description: Gaetano is a software engineer focused on pragmatic developer
+  tooling, mobile application workflows, and AI-assisted coding systems. He
+  writes about reproducible engineering processes that make agent output easier
+  to test, review, and trust.
+Author Image: [https://avatars.githubusercontent.com/skrtdev?size=512]
+Author LinkedIn:
+Author Twitter:
+Company Name:
+Company Description:
+Company Logo Dark:
+Company Logo White:
diff --git a/definitions/20260604_definition_accessibility_regression.md b/definitions/20260604_definition_accessibility_regression.md
new file mode 100644
index 00000000..68fda63f
--- /dev/null
+++ b/definitions/20260604_definition_accessibility_regression.md
@@ -0,0 +1,29 @@
+---
+title: 'Accessibility Regression'
+description:
+  'An accessibility regression is a code change that makes a product harder to
+  use with assistive technology, keyboard navigation, or other access needs.'
+date: 2026-06-04
+author: 'Gaetano'
+---
+
+# Accessibility Regression
+
+## Definition
+
+An accessibility regression is a change that reduces a product's usability for
+people who rely on assistive technology, keyboard navigation, captions,
+contrast, predictable focus, or semantic structure. The feature may still look
+correct visually while becoming harder or impossible to use for some users.
+
+## Context and Usage
+
+Accessibility regressions often appear in small interface edits. A developer
+can replace a labeled input with placeholder text, turn a text button into an
+icon-only button without an accessible name, remove `aria-live` from a dynamic
+status message, or break keyboard focus order while reorganizing a form.
+
+For example, changing a reset button from `Reset filters` to a visual `x`
+without `aria-label="Reset filters"` can remove the button's accessible name.
+The visual design still appears clear to sighted users, but screen reader users
+hear an ambiguous control.
diff --git a/guides/20260604_guide_rehearse_accessibility_regression_triage_in_daytona.md b/guides/20260604_guide_rehearse_accessibility_regression_triage_in_daytona.md
new file mode 100644
index 00000000..c5db6413
--- /dev/null
+++ b/guides/20260604_guide_rehearse_accessibility_regression_triage_in_daytona.md
@@ -0,0 +1,364 @@
+---
+title: 'Rehearse Accessibility Triage in Daytona'
+description:
+  'Use Omni Engineer and Claude Engineer in a Daytona sandbox to reproduce,
+  repair, review, and verify an accessibility regression.'
+date: 2026-06-04
+author: 'Gaetano'
+tags: ['AI', 'accessibility', 'daytona', 'devcontainer']
+---
+
+# Rehearse Accessibility Triage in Daytona
+
+# Introduction
+
+Accessibility bugs are easy to miss when a change still looks correct in a
+browser. A filter form can keep its visual layout while losing a label, an
+icon-only reset button can lose its accessible name, and a dynamic status line
+can stop announcing updates to assistive technology. That kind of
+[accessibility regression](../definitions/20260604_definition_accessibility_regression.md)
+often survives casual review because the interface still appears usable to the
+developer who changed it.
+
+AI engineering tools can help, but they need a tight workflow. In this guide,
+[Omni Engineer](https://github.com/Doriandarko/omni-engineer) acts as the
+repair agent, [Claude Engineer](https://github.com/Doriandarko/claude-engineer)
+acts as the accessibility reviewer, and a [Daytona
+sandbox](https://www.daytona.io/docs/en/sandboxes/) keeps the exercise
+isolated. The goal is not to ask an agent to "make accessibility better." The
+goal is to give two agents one reproducible failure, one focused patch, one
+review pass, and one executable test suite.
+
+You will use a small Flask demo project with a deliberately broken branch:
+
+- [Daytona Accessibility Regression Demo](https://github.com/skrtdev/daytona-a11y-regression-demo)
+
+The `regression` branch fails tests for four common issues: a search input
+without a real label, a reset button without an accessible name, a status
+message without a polite live region, and filter buttons without grouped
+pressed state. The `main` branch is the fixed reference implementation.
+
+## TL;DR
+
+- Add secret-free Dev Container definitions to Omni Engineer and Claude
+  Engineer so each repository has a portable Python workspace.
+- Create one Daytona sandbox with API keys supplied as environment variables.
+- Clone the demo project and check out its `regression` branch.
+- Ask Omni Engineer to make the smallest semantic markup fix that turns the
+  failing tests green.
+- Ask Claude Engineer to review the exact diff for keyboard, screen reader, and
+  regression-test gaps.
+- Run `pytest -q` and `git diff --check` before treating the repair as ready.
+
+## Step 1: Prepare the Agent Workspaces
+
+The companion Dev Container contributions for this workflow are:
+
+| Repository | Purpose | Pull Request |
+| --- | --- | --- |
+| Omni Engineer | Python 3.11 workspace, dependency install, and `OPENROUTER_API_KEY` passthrough | [Doriandarko/omni-engineer#45](https://github.com/Doriandarko/omni-engineer/pull/45) |
+| Claude Engineer | Python 3.11 workspace, dependency install, `ANTHROPIC_API_KEY` passthrough, optional `E2B_API_KEY`, and port `5000` forwarding | [Doriandarko/claude-engineer#270](https://github.com/Doriandarko/claude-engineer/pull/270) |
+
+The files are intentionally small. They install each repository's existing
+`requirements.txt` and read keys from the local environment:
+
+```json
+{
+  "image": "mcr.microsoft.com/devcontainers/python:3.11",
+  "postCreateCommand": "python -m pip install --upgrade pip && python -m pip install -r requirements.txt",
+  "remoteUser": "vscode"
+}
+```
+
+The full Omni Engineer version passes `OPENROUTER_API_KEY`. The Claude Engineer
+version passes `ANTHROPIC_API_KEY`, optionally passes `E2B_API_KEY`, and
+forwards port `5000` for the web UI. Neither file includes a `.env` file or
+hard-codes a secret.
+
+Daytona's current CLI creates and executes commands in sandboxes. The
+walkthrough below uses the documented `daytona create`, `daytona exec`, and
+`daytona ssh` flow instead of assuming that Daytona automatically interprets a
+repository's `devcontainer.json`. The Dev Container files are still useful
+because they make each agent repository portable for editors and other
+compatible development environments.
+
+## Step 2: Create a Daytona Sandbox
+
+Install and authenticate the Daytona CLI before starting. The CLI reference
+documents `daytona create --env` for environment variables and `daytona exec
+--cwd` for command execution in a selected working directory.
+
+Export your keys locally:
+
+```bash
+export OPENROUTER_API_KEY="your-openrouter-key"
+export ANTHROPIC_API_KEY="your-anthropic-key"
+```
+
+Create an isolated sandbox:
+
+```bash
+daytona create \
+  --name a11y-triage-lab \
+  --env OPENROUTER_API_KEY="$OPENROUTER_API_KEY" \
+  --env ANTHROPIC_API_KEY="$ANTHROPIC_API_KEY"
+```
+
+> **Note:** Keep real keys outside Git. If you create the sandbox before
+> exporting the variables, recreate the sandbox with the correct `--env` flags
+> rather than committing a local `.env` file.
+
+Clone the two agents and the demo project:
+
+```bash
+daytona exec a11y-triage-lab \
+  -- git clone https://github.com/skrtdev/omni-engineer.git /home/daytona/omni-engineer
+
+daytona exec a11y-triage-lab \
+  -- git clone https://github.com/skrtdev/claude-engineer.git /home/daytona/claude-engineer
+
+daytona exec a11y-triage-lab \
+  -- git clone https://github.com/skrtdev/daytona-a11y-regression-demo.git /home/daytona/daytona-a11y-regression-demo
+```
+
+Check out the intentionally broken branch:
+
+```bash
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- git checkout regression
+```
+
+Create a separate virtual environment for each repository:
+
+```bash
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/omni-engineer \
+  -- sh -lc "python -m venv .venv && .venv/bin/python -m pip install -r requirements.txt"
+
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/claude-engineer \
+  -- sh -lc "python -m venv .venv && .venv/bin/python -m pip install -r requirements.txt"
+
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- sh -lc "python -m venv .venv && .venv/bin/python -m pip install -r requirements.txt"
+```
+
+![Accessibility regression triage flow](assets/20260604_guide_rehearse_accessibility_regression_triage_in_daytona_img1.svg)
+
+The single sandbox keeps the work honest. Omni Engineer sees the failing files,
+Claude Engineer reviews the same patch, and the tests run in the same
+filesystem instead of in a pasted transcript.
+
+## Step 3: Reproduce the Regression First
+
+Run the tests before asking any agent to change code:
+
+```bash
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- .venv/bin/python -m pytest -q
+```
+
+The `regression` branch should fail four tests:
+
+| Failing Check | What It Means |
+| --- | --- |
+| `test_search_input_has_visible_label` | The search input is visually described but not associated with a real `<label>` |
+| `test_reset_button_has_accessible_name` | The icon-only reset button has no accessible name |
+| `test_filter_status_is_polite_live_region` | The status text will not be announced when it changes |
+| `test_filter_controls_are_grouped_with_pressed_state` | The filter buttons do not expose grouped state to assistive technology |
+
+This step matters. Reproduction gives the agents a measurable problem and gives
+you a baseline. Without it, an agent may produce a broad rewrite that sounds
+helpful but does not prove anything about the original bug.
+
+## Step 4: Ask Omni Engineer for the Smallest Fix
+
+Open a shell in the sandbox:
+
+```bash
+daytona ssh a11y-triage-lab
+```
+
+Start Omni Engineer:
+
+```bash
+cd /home/daytona/omni-engineer
+.venv/bin/python main.py
+```
+
+Add the demo files to context:
+
+```text
+/add ../daytona-a11y-regression-demo/templates/index.html
+/add ../daytona-a11y-regression-demo/tests/test_accessibility.py
+```
+
+Then give Omni Engineer a narrow repair request:
+
+```text
+Repair the accessibility regression in
+/home/daytona/daytona-a11y-regression-demo.
+
+Acceptance criteria:
+1. Keep the existing Flask route and visible UI copy.
+2. Associate the search input with a real visible label.
+3. Give the icon-only reset button an accessible name.
+4. Restore a polite live region for the filter status message.
+5. Group the issue-state filter buttons and expose aria-pressed state.
+6. Make the existing pytest accessibility tests pass.
+7. Show the diff before treating the fix as complete.
+
+Only change templates/index.html unless the tests prove another file must
+change.
+```
+
+The key phrase is "only change `templates/index.html`." Accessibility fixes can
+tempt an agent into redesigning the component, renaming tests, or deleting
+assertions. A bounded prompt keeps the repair focused on semantics.
+
+After Omni Engineer proposes the patch, run the tests:
+
+```bash
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- .venv/bin/python -m pytest -q
+```
+
+The expected passing state is:
+
+```text
+4 passed
+```
+
+If the tests still fail, feed the exact failure back into Omni Engineer and ask
+for the smallest follow-up patch.
+
+## Step 5: Use Claude Engineer as the Reviewer
+
+Once the tests pass, use Claude Engineer to review the exact diff. Start the
+CLI or web interface. The CLI keeps the workflow terminal-first:
+
+```bash
+cd /home/daytona/claude-engineer
+.venv/bin/python ce3.py
+```
+
+Ask for an accessibility review, not a rewrite:
+
+```text
+Review the accessibility regression fix in
+/home/daytona/daytona-a11y-regression-demo.
+
+Inspect templates/index.html, tests/test_accessibility.py, and the git diff.
+Focus on whether the fix preserves keyboard access, visible labels,
+accessible names, live-region behavior, grouped control state, and test
+coverage. Do not rewrite the UI unless you find a concrete regression.
+If there is a gap, explain it first and propose the smallest patch.
+```
+
+This is where the two-agent pattern pays off. Omni Engineer is biased toward
+making the failure disappear. Claude Engineer is biased toward asking whether
+the disappearance is meaningful. That difference catches issues such as:
+
+- A hidden label that passes a test but makes the visual UI less clear.
+- A reset button that has an accessible name but removes the visible affordance.
+- A live region that exists but no longer contains the status message.
+- Filter buttons that expose `aria-pressed` but no longer sit inside a named
+  group.
+- Tests that were changed to match the broken implementation.
+
+If Claude Engineer recommends a patch, apply it and rerun the same tests. Do
+not accept a review comment as proof until the repository verifies it.
+
+## Step 6: Confirm and Preserve the Evidence
+
+Run the final checks:
+
+```bash
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- git diff --check
+
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- .venv/bin/python -m pytest -q
+```
+
+Inspect the final markup:
+
+```bash
+daytona exec a11y-triage-lab \
+  --cwd /home/daytona/daytona-a11y-regression-demo \
+  -- sed -n '1,120p' templates/index.html
+```
+
+The fixed markup should include:
+
+```html
+<label for="issue-search">Search issues</label>
+<input id="issue-search" name="q" type="search" autocomplete="off">
+
+<div role="group" aria-label="Issue state filters">
+  <button type="button" aria-pressed="true">All</button>
+</div>
+
+<button type="reset" aria-label="Reset filters">x</button>
+<p id="filter-status" role="status" aria-live="polite">Showing all issues.</p>
+```
+
+You do not need to ship this exact formatting. You do need to preserve the
+semantics: a real label, an accessible name for the icon-only control, a named
+group with pressed state, and a polite live region.
+
+## Troubleshooting
+
+**Problem: The agent changes the tests instead of the markup.**
+
+Ask it to revert the test edits and repair only `templates/index.html`. The
+tests are the contract for this exercise.
+
+**Problem: The agent adds `aria-label` to the search input instead of restoring
+the visible label.**
+
+Reject that patch for this specific UI. A visible label helps more users and is
+already part of the intended design.
+
+**Problem: Claude Engineer suggests a complete component rewrite.**
+
+Ask for the smallest patch that resolves a concrete accessibility or test
+coverage gap. Broad rewrites make regressions harder to review.
+
+**Problem: Daytona commands cannot find the repository path.**
+
+Run `daytona exec a11y-triage-lab -- ls /home/daytona` and confirm that the
+clone commands used the same directory names as the guide.
+
+**Problem: An API key is unavailable inside the sandbox.**
+
+Recreate the sandbox with `--env OPENROUTER_API_KEY=...` and
+`--env ANTHROPIC_API_KEY=...`. Do not commit real keys into either agent repo.
+
+## Conclusion
+
+Accessibility triage works best when the problem is reproducible and the fix is
+small. Daytona gives the agents an isolated workspace. Omni Engineer repairs
+the failing semantics. Claude Engineer reviews the exact patch. The test suite
+proves whether the label, button name, live region, and grouped filter state
+survived the change.
+
+That pattern scales beyond this demo. Any agent-assisted accessibility task can
+use the same shape: reproduce first, patch narrowly, review separately, and
+verify with executable checks before trusting the result.
+
+## References
+
+- [Daytona Sandboxes](https://www.daytona.io/docs/en/sandboxes/)
+- [Daytona CLI Reference](https://www.daytona.io/docs/en/tools/cli/)
+- [Daytona Getting Started](https://www.daytona.io/docs/en/getting-started/)
+- [Dev Container Specification](https://containers.dev/)
+- [Omni Engineer Repository](https://github.com/Doriandarko/omni-engineer)
+- [Claude Engineer Repository](https://github.com/Doriandarko/claude-engineer)
+- [Daytona Accessibility Regression Demo](https://github.com/skrtdev/daytona-a11y-regression-demo)
diff --git a/guides/assets/20260604_guide_rehearse_accessibility_regression_triage_in_daytona_img1.svg b/guides/assets/20260604_guide_rehearse_accessibility_regression_triage_in_daytona_img1.svg
new file mode 100644
index 00000000..58c25f9a
--- /dev/null
+++ b/guides/assets/20260604_guide_rehearse_accessibility_regression_triage_in_daytona_img1.svg
@@ -0,0 +1,35 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="1200" height="675" viewBox="0 0 1200 675" role="img" aria-labelledby="title desc">
+  <title id="title">Accessibility regression triage in a Daytona sandbox</title>
+  <desc id="desc">Omni Engineer fixes a failing accessibility demo, Claude Engineer reviews the exact patch, and pytest verifies the result inside one Daytona sandbox.</desc>
+  <rect width="1200" height="675" fill="#f7fafc"/>
+  <rect x="52" y="54" width="1096" height="567" rx="14" fill="#ffffff" stroke="#111827" stroke-width="3"/>
+  <text x="92" y="112" fill="#111827" font-family="Arial, sans-serif" font-size="34" font-weight="700">Daytona sandbox: a11y-triage-lab</text>
+  <text x="92" y="151" fill="#4b5563" font-family="Arial, sans-serif" font-size="20">One isolated workspace for reproduction, agent repair, review, and verification</text>
+
+  <rect x="95" y="228" width="250" height="145" rx="10" fill="#dbeafe" stroke="#1d4ed8" stroke-width="3"/>
+  <text x="220" y="280" text-anchor="middle" fill="#1e3a8a" font-family="Arial, sans-serif" font-size="26" font-weight="700">Regression branch</text>
+  <text x="220" y="318" text-anchor="middle" fill="#1e40af" font-family="Arial, sans-serif" font-size="18">Missing label, live</text>
+  <text x="220" y="344" text-anchor="middle" fill="#1e40af" font-family="Arial, sans-serif" font-size="18">region, and button name</text>
+
+  <rect x="475" y="228" width="250" height="145" rx="10" fill="#dcfce7" stroke="#15803d" stroke-width="3"/>
+  <text x="600" y="280" text-anchor="middle" fill="#14532d" font-family="Arial, sans-serif" font-size="26" font-weight="700">Omni Engineer</text>
+  <text x="600" y="318" text-anchor="middle" fill="#166534" font-family="Arial, sans-serif" font-size="18">Applies the smallest</text>
+  <text x="600" y="344" text-anchor="middle" fill="#166534" font-family="Arial, sans-serif" font-size="18">semantic markup fix</text>
+
+  <rect x="855" y="228" width="250" height="145" rx="10" fill="#fef3c7" stroke="#b45309" stroke-width="3"/>
+  <text x="980" y="280" text-anchor="middle" fill="#78350f" font-family="Arial, sans-serif" font-size="26" font-weight="700">Claude Engineer</text>
+  <text x="980" y="318" text-anchor="middle" fill="#92400e" font-family="Arial, sans-serif" font-size="18">Reviews keyboard and</text>
+  <text x="980" y="344" text-anchor="middle" fill="#92400e" font-family="Arial, sans-serif" font-size="18">assistive tech risks</text>
+
+  <line x1="345" y1="300" x2="455" y2="300" stroke="#111827" stroke-width="4"/>
+  <polygon points="455,300 438,290 438,310" fill="#111827"/>
+  <line x1="725" y1="300" x2="835" y2="300" stroke="#111827" stroke-width="4"/>
+  <polygon points="835,300 818,290 818,310" fill="#111827"/>
+
+  <rect x="366" y="458" width="468" height="94" rx="10" fill="#fce7f3" stroke="#be185d" stroke-width="3"/>
+  <text x="600" y="500" text-anchor="middle" fill="#831843" font-family="Arial, sans-serif" font-size="26" font-weight="700">pytest accessibility checks</text>
+  <text x="600" y="532" text-anchor="middle" fill="#9d174d" font-family="Arial, sans-serif" font-size="18">Labels, accessible names, live regions, and pressed state</text>
+
+  <line x1="600" y1="373" x2="600" y2="438" stroke="#111827" stroke-width="4"/>
+  <polygon points="600,438 590,421 610,421" fill="#111827"/>
+</svg>