docs(#6163): add staging warning to go-code.md coverage procedure

[fullsend-ai-coder]

Summary

Add a staging warning to the "Verifying patch coverage locally" procedure in docs/contributing/go-code.md step 1, matching the equivalent warning already present in skills/check-patch-coverage/SKILL.md step 1. Without this note, an agent following go-code.md would not know to git add new files before running git diff --name-only, causing the coverage check to silently skip unstaged new files.

Changes

• Added "Stage new files first (git add)" note to step 1 of the coverage procedure in docs/contributing/go-code.md
• Wording is consistent with the existing warning in skills/check-patch-coverage/SKILL.md

Testing

• Verified both docs/contributing/go-code.md and skills/check-patch-coverage/SKILL.md contain equivalent staging warnings
• No Go code changed — no tests or linters applicable

Checklist

• PR title follows Conventional Commits (correct type, ! for breaking changes)

---

Closes #6163

Post-script verification

• Branch is not main/master (agent/6163-add-staging-warning)
• Secret scan passed (gitleaks — 3c7f5d034d3f9ace8efd3608a6322903da907809..HEAD)
• PR body secret scan passed (gitleaks — no-git)

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 6:01 AM UTC · Completed 6:09 AM UTC

Commit: 75b4b2a · View workflow run →

[github-actions]

Site preview

Preview: https://a444f89c-site.fullsend-ai.workers.dev

Commit: 75b4b2aa15f4495e68c68d6202eeeb9859daa185

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-review]

Looks good to me

---

Labels: PR modifies documentation under docs/contributing/

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 8:49 AM UTC · Completed 9:01 AM UTC

Commit: 75b4b2a · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $4.77

[fullsend-ai-retro]

Retro: PR #6164 — docs(#6163): add staging warning to go-code.md coverage procedure

Verdict: Clean end-to-end autonomous workflow. No new proposals — all findings map to existing open issues.

Timeline

1. Issue filed (Aug 13, 05:50 UTC): The retro agent identified a documentation inconsistency from PR #6138 — a staging warning was added to skills/check-patch-coverage/SKILL.md but not to docs/contributing/go-code.md. Filed as #6163.
2. Triage (05:50–05:54, ~3.5 min, $0.50): Correctly classified as low-priority documentation fix. Verified the inconsistency by reading both files. Applied appropriate labels and marked ready-to-code.
3. Code (05:54–05:59, ~6 min): Made the correct 4-line edit on the first attempt with zero dead ends. Single commit, clean branch.
4. CI (05:59–06:03): All checks passed.
5. Review (06:00–06:10, ~10 min, $2.28): Dispatched 5 sub-agents (correctness, style-conventions, intent-coherence, docs-currency, challenger). Correctly approved with "Looks good to me." Docs-currency sub-agent efficiently self-skipped (0 tool calls).
6. Human review (Aug 26, 08:42): rh-hemartin approved. PR merged at 08:48.

Workflow quality

• Rework: Zero iterations. Code agent got the change right on the first attempt.
• Review accuracy: Agent review was correct — the change faithfully ports existing staging-warning text from SKILL.md to go-code.md. No false positives emitted.
• Token efficiency: Code agent was lean (14.5K output tokens, 33 tool calls, 1 edit). Review agent was somewhat over-provisioned for the change size ($2.28 for a 4-line docs edit) — the style-conventions sub-agent made 7 tool calls including reading unrelated docs (shell-scripting.md, adrs.md), and SKILL.md was read redundantly by 4 of 5 sub-agents.
• Time to resolution: 13 days from PR open to merge. The bottleneck was human review latency, not agent performance.

Evidence for existing issues (skipped as proposals)

• Review fast-path for trivial docs PRs: This PR ($2.28, 10 min, 5 sub-agents for a 4-line, 1-file docs change) provides additional evidence for fullsend-ai/fullsend#1900 (fast path for docs/prompt-only changes) and fullsend-ai/agents#373 (skip irrelevant review dimensions for docs-only PRs).
• Review cost proportionality: The style-conventions sub-agent's exploration of unrelated contributing docs supports fullsend-ai/agents#439 (scale review detail to change complexity).
• Redundant sub-agent file reads: Four sub-agents independently reading SKILL.md supports the general context-packaging improvements tracked in fullsend-ai/fullsend#2836 (exploration budgets for sub-agents).
• Retro-originated PR chain: This PR was itself filed by the retro agent from PR docs(#2810): add patch coverage verification guidance for agents #6138, and now triggers another retro run — relevant to fullsend-ai/fullsend#5319 (detect retro-originated PR chains).

What went well

This is a success story for the retro → triage → code → review pipeline. The retro agent identified a real documentation inconsistency, the triage agent correctly scoped it, the code agent resolved it on the first attempt, and the review agent correctly approved. The full autonomous chain from issue detection to merge-ready PR completed in under 20 minutes with zero rework.