Improve subagent profile examples with concrete models and QA profile

[Waishnav]

Summary

Polishes the built-in examples/agents/* subagent profiles so they read like actionable playbooks instead of generic “you are a worker” prompts. Adds explicit model / thinking frontmatter where helpful and introduces a dedicated manual QA example.

What changed

New profile

• codex-qa-tester — read-only manual QA pass: acceptance-criteria checklist, browser/workflow verification, reproduction steps, and untested risks.

Updated profiles (7 files)

Profile	Highlights
claude-implementer	Clearer implementation rules; report drops files_changed in favor of focused fields
codex-explorer	gpt-5.4-mini + thinking: high; evidence-first exploration guidance
codex-worker	Tighter scope rules; notes instead of follow_up_needed in report
copilot-reviewer	Severity-ordered findings, test gaps, residual risk
cursor-agent-worker	composer-2.5-fast; UI/responsive verification emphasis
opencode-explorer	opencode/deepseek-v4-flash-free + thinking: high; search-then-read workflow
pi-reviewer	openai-codex/gpt-5.5 + thinking: high; actionable risk focus

Themes across edits

• Descriptions — task-oriented (what to use the profile for), not provider marketing copy.
• Body copy — bullet “working rules” instead of role-play framing.
• Reports — standardized Report: blocks with fields aligned to each role (explorer vs implementer vs reviewer vs QA).
• Models — concrete defaults in YAML so copy-paste examples work out of the box with current provider options.

Commits

• 1033109 Improve subagent profile examples
• c823cad Add concrete models to subagent examples

Testing

• Documentation-only / example markdown; no runtime code changes on this branch.

Related context

Builds on merged subagent provider + thinking profile support (39710ce … 74b8c22 on main). This PR is examples-only.

Summary by CodeRabbit

• Documentation

  • Updated several example agent profiles with clearer, more focused guidance for implementation, review, exploration, and QA workflows.
  • Simplified and standardized report templates across examples for more consistent output.
  • Added a new QA tester example for browser-based workflow verification.

• Chores

  • Refreshed agent metadata such as descriptions, model selections, and thinking settings.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR revises eight example Devspace agent profile markdown files under examples/agents/. Frontmatter (description, model, thinking level) and instruction text are updated across implementer, explorer, worker, and reviewer profiles, with each final report template restructured. A new codex-qa-tester.md profile is added for manual QA testing.

Changes

Devspace Agent Profile Documentation

Layer / File(s)	Summary
Claude implementer profile examples/agents/claude-implementer.md	Description, working rules, and report template (summary, tests_run, blockers, risks) are revised.
Codex worker profile examples/agents/codex-worker.md	Description, rules, and report schema (summary, tests_run, blockers, notes) are revised.
Cursor agent worker profile examples/agents/cursor-agent-worker.md	Description, model (composer-2.5-fast), instructions, and report template (summary, verification, blockers, open_questions) are updated.
Codex explorer profile examples/agents/codex-explorer.md	Description, model (gpt-5.4-mini), thinking: high, investigation instructions, and report (answer, evidence, relevant_files) are updated.
Opencode explorer profile examples/agents/opencode-explorer.md	Description, model, thinking: high, instructions, and report format are updated similarly to the Codex explorer.
Copilot reviewer profile examples/agents/copilot-reviewer.md	Description and review instructions/report (findings, evidence, test_gaps, residual_risk) are updated.
Pi reviewer profile examples/agents/pi-reviewer.md	Description, model (openai-codex/gpt-5.5), thinking: high, and report (findings, evidence, risk_level) are updated.
New QA tester profile examples/agents/codex-qa-tester.md	A new agent config is added with metadata and a browser-based QA testing workflow and report template.

Estimated code review effort: 2 (Simple) | ~12 minutes

Estimated code review effort: 2 (Simple) | ~12 minutes

Poem

Eight little profiles, hopping in a row,
New rules, new reports, all set to go.
One fresh QA bunny joins the burrow crew,
Testing every pathway, old and new.
Thump thump — docs approved, the warren's true! 🐇📋

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title matches the main change set: updated subagent profile examples, added concrete models, and introduced a QA profile.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/subagent-profile-examples

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@examples/agents/claude-implementer.md`:
- Around line 22-30: The report template includes an extra follow_up_needed
field that breaks the standardized summary/tests_run/blockers/risks shape.
Update the report block in the Claude implementer example to remove
follow_up_needed and keep the template aligned with the other report examples,
preserving only the four expected keys.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 993d745e-4cbd-435c-a396-926ce1b19c64

📥 Commits

Reviewing files that changed from the base of the PR and between 11b1e8e and c823cad.

📒 Files selected for processing (8)

• examples/agents/claude-implementer.md
• examples/agents/codex-explorer.md
• examples/agents/codex-qa-tester.md
• examples/agents/codex-worker.md
• examples/agents/copilot-reviewer.md
• examples/agents/cursor-agent-worker.md
• examples/agents/opencode-explorer.md
• examples/agents/pi-reviewer.md

[coderabbitai]

📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win

Drop the extra follow_up_needed key.

The updated report template is meant to standardize on summary/tests_run/blockers/risks, but this extra field keeps the shape inconsistent with the rest of the examples. As per the PR objective to standardize report blocks, keep the schema aligned here.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@examples/agents/claude-implementer.md` around lines 22 - 30, The report
template includes an extra follow_up_needed field that breaks the standardized
summary/tests_run/blockers/risks shape. Update the report block in the Claude
implementer example to remove follow_up_needed and keep the template aligned
with the other report examples, preserving only the four expected keys.