docs(#6941): add SandboxCodexConfig to CLI internals guide

[fullsend-ai-coder]

Summary

Add missing SandboxCodexConfig constant to the Sandbox Constants code block in docs/guides/dev/cli-internals.md. PR #6923 added this constant to internal/sandbox/sandbox.go but did not update the documentation, leaving developers without a reference to the codex config path.

Changes

• Add SandboxCodexConfig = "/sandbox/codex-config" between SandboxClaudeConfig and SandboxPiConfig in the Sandbox Constants code block, matching the source ordering in internal/sandbox/sandbox.go

Testing

• Verified all five constants now appear in the docs code block in the same order as the source file
• Pre-commit hooks pass (trailing-whitespace, mixed-line-ending, check-merge-conflict, lint-docs-links, lint-md-links/lychee)
• Secret scan passes

---

Closes #6941

Post-script verification

• Branch is not main/master (agent/6941-add-sandbox-codex-config-docs)
• Secret scan passed (gitleaks — 0adea5decc57f61e234999e84008bedc96e18d31..HEAD)
• PR body secret scan passed (gitleaks — no-git)

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 11:46 PM UTC · Completed 11:57 PM UTC

Commit: 50f1a48 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.51

[github-actions]

Site preview

Preview: https://d1decbec-site.fullsend-ai.workers.dev

Commit: 50f1a4878432b0330d6b71a0774b38d864636da9

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-review]

Risk Assessment: low (1/5)

Details

Minimal-risk docs-only PR: a single-line addition to cli-internals.md by a known bot, fixing a documentation gap flagged in a priority/low chore issue. No protected paths, security-sensitive files, CI workflows, or dependencies touched.

[fullsend-ai-review]

Looks good to me

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 10:00 AM UTC · Completed 10:10 AM UTC

Commit: 50f1a48 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.99

[fullsend-ai-retro]

Retro: PR #6942 — docs(#6941): add SandboxCodexConfig to CLI internals guide

Timeline

PR #6923 added SandboxCodexConfig to internal/sandbox/sandbox.go but did not update the documentation. The review agent flagged this docs-currency gap twice during PR #6923's review, but the human reviewer approved and merged without addressing it. The retro agent on PR #6923 then filed issue #6941 with a high-confidence, validated proposal.

Step	Agent	Time (UTC)	Cost	Run
Issue filed	retro	2026-09-02 23:32	—	(from PR #6923 retro)
Triage	triage	23:33–23:38	$0.59	33695580618
Code	code	23:38–23:44	$1.30	33695984854
Review	review	23:44–23:57	$2.51	33696470652
Human approval	rh-hemartin	2026-09-04 09:53	—	—
Merged	—	09:58	—	—
Retro dispatched	retro	09:59	—	33860994499

Total agent cost: $4.40 for a single-line docs fix. Agents repo: fullsend-ai/agents@daa53e09.

What went well

• Zero rework. The code agent added the correct line in the correct position on the first try. No fix agent cycles needed.
• Correct risk assessment. The review agent rated the PR as low risk (1/5) and approved with no findings — appropriate for a one-line docs change.
• Full agent-human agreement. The review agent and human reviewer (rh-hemartin) both approved without requesting changes. No delta between agent and human review.
• Fast automated pipeline. From issue creation to PR opened and reviewed: ~25 minutes. The 34-hour gap to merge was human review latency, which is normal for async workflows.

Evidence for existing open issues

• #6891 (pre-flight scope heuristic): The review agent spent $2.51 at high effort with claude-opus-4-6 on a 1-line docs-only bot-authored PR. This is 57% of the total agent cost for the workflow. A pre-flight scope check would have routed this to a lighter model/effort. See also #6666 (stale sonnet mapping) and #6964 (model availability validation), which may compound the cost.
• #1475 / #3841 (fast-track triage for retro-filed issues): The triage agent spent $0.59 to verify what the retro agent had already verified and documented with file paths, line numbers, and validation criteria. The triage comment restated the same gap with no novel information. For high-confidence retro-originated issues, triage is redundant.
• #5319 (retro-on-retro-fix dispatch): This retro run analyzes PR docs(#6941): add SandboxCodexConfig to CLI internals guide #6942, which implements retro-originated issue Fix docs-currency gap: SandboxCodexConfig missing from CLI internals guide #6941. Running retro on a trivially successful implementation of a retro proposal is the exact self-referential pattern this issue describes.
• #6210 / #5289 (autonomy readiness): The review agent and human reviewer were in complete agreement on this docs-only, single-line, bot-authored PR — zero delta. This is another data point supporting automated approval for low-risk docs-only bot PRs.

Conclusion

The workflow executed correctly end-to-end with no failures or rework. All improvement opportunities identified are covered by existing open issues. No new proposals filed.