[codex] Clarify SDD code review overlap

[Derekko-web]

What problem are you trying to solve?

Fixes #1481. The current requesting-code-review wording can be read as requiring a second standalone review after SDD has already run its built-in per-task code quality review. That can waste subagent calls and create duplicate feedback for the same diff.

What does this PR change?

This clarifies that SDD invokes the requesting-code-review mechanism through code-quality-reviewer-prompt.md, and that the built-in SDD code quality review satisfies this skill for the reviewed diff. It also narrows the major-feature trigger to major features completed outside SDD.

Is this change appropriate for the core library?

Yes. This affects the relationship between two core workflow skills and is not project-specific.

What alternatives did you consider?

One alternative was to remove the SDD mention from requesting-code-review entirely, but that would hide the fact that SDD still requires per-task review. Another was to leave the wording as-is and rely on agents to infer the relationship, but the current mandatory wording can reasonably be interpreted as a separate post-SDD trigger.

Does this PR contain multiple unrelated changes?

No. It changes one file and only clarifies the SDD/requesting-code-review boundary.

Existing PRs

• I have reviewed all open AND closed PRs for duplicates or prior art
• Related PRs: no direct duplicate found for Clarify that SDD code quality review satisfies requesting-code-review #1481. Broad workflow PRs such as fix(skills): promote worktree step into executor process flows #1297 and feat: add lifecycle extension system for user-defined workflow hooks #1224 touch nearby SDD docs, but do not address this specific duplicate-review clarification.

Environment tested

Harness (e.g. Claude Code, Cursor)	Harness version	Model	Model version/ID
Codex desktop	current session	GPT-5	session default

New harness support (required if this PR adds a new harness)

Not applicable. This PR does not add a new harness.

Clean-session transcript for "Let's make a react todo list"

Not applicable.

Evaluation

• Initial prompt: "do this for me" after identifying a small first-contributor docs issue. writing-skills: missing guidance on quoting description values in YAML frontmatter #955 was skipped because open PRs Docs: Require quoted YAML description in SKILL.md and clarify description guidance #979 and docs(writing-skills): add YAML frontmatter quoting guidance #1028 already cover it, so this PR targets Clarify that SDD code quality review satisfies requesting-code-review #1481 instead.
• Eval sessions run after the change: 0 adversarial agent sessions; this draft needs human review before it is ready.
• Static checks run:
  • git diff --check
  • targeted rg assertion confirming the old ambiguous SDD wording is gone and the new satisfaction wording appears exactly once

I also attempted tests/claude-code/run-skill-tests.sh --test test-subagent-driven-development.sh; the runner depends on GNU timeout, which is not installed in this macOS environment. A temporary Bash shim got the runner started but left the wrapper stuck with no active claude child process, so I terminated that attempt rather than report it as a pass.

Rigor

• If this is a skills change: I used superpowers:writing-skills and completed adversarial pressure testing (paste results below)
• This change was tested adversarially, not just on the happy path
• I did not modify carefully-tuned content (Red Flags table, rationalizations, "human partner" language) without extensive evals showing the change is an improvement

This is opened as a draft because the project template requires human review and stronger skill-change evaluation before the PR is ready for maintainer review.

Human review

• A human has reviewed the COMPLETE proposed diff before submission