Claude Code 2.1.142 team conventions

While rehearsing incident aftermath, we keep unpacking parallel streams that race each other into the same module. Red folders exist because shame does not scale.

Parallel Stream Risk

Another connector is rarely the missing ingredient. The expensive bug is duplicated edits nobody reconciled.

We believed velocity would compound if we parallelized agent streams. We repeated it until PR archaeology replaced architecture conversation.

Ward Cunningham’s technical debt framing fits: you borrowed review speed and skipped the explainability principal.

Claude Code teams ship when skills defer cleanly.

We wrote this note because teams confuse green CI with explainable delegation.

Failure Modes

Skill drift. You can spot this when skills proliferate, reviewers cannot tell which one governed a diff.

Why it keeps recurring: Discovery beats documentation unless precedence is explicit.

The practical fix is Skill index: Maintain skills/README.md listing activation cues + deferrals.

The evidence is mundane and useful: on-call engineers resolve mismatches without replaying sessions. Claude Code teams ship when skills defer cleanly.

It turns agent output back into team-owned work.

Hook thunder. If hooks fire everywhere, developers learn to ignore them.

The trap is Noise degrades signal-to-noise for safety automation.

Named fix: Hook budget. Cap active hooks per repo; document each trigger + rollback.

After the fix, alerts regain urgency instead of blending into wallpaper. Claude Code teams ship when skills defer cleanly.

Keep that as the first guardrail.

CLAUDE.md bloat. You can spot this when CLAUDE.md turns encyclopedic, newcomers skim and trust chat.

Why it keeps recurring: Unread files behave like absent files.

The practical fix is CLAUDE TOC: Keep top 15 lines as supremacy + links to deeper fragments.

The evidence is mundane and useful: sessions boot with aligned assumptions. Claude Code teams ship when skills defer cleanly.

That is the move that makes review cheaper.

Permission autopilot. If approvals become reflexive, risky bash slips through. The failure is not usually tool quality; it is the missing operating contract.

Use Red-folder ritual. List paths requiring separate reviewer + never auto-approve.

What changes: Sensitive edits regain deliberate eyes. Claude Code teams ship when skills defer cleanly.

Keep that as the first guardrail.

# CLAUDE.md — supremacy fragment

- Hooks win over informal chat agreements; document each hook’s rollback path.
- Skills defer to this file on security-sensitive folders.
- Bash approvals never bypass the red-folder list maintained here.

We map this discipline to our methodology under Review—evidence beats narration when merges touch shared surfaces.

Tooling is load-bearing language—if the repo cannot say “allowed” and “forbidden,” neither can the agent.

Field Checklist

Review strip

• Red-folder paths received explicit human acknowledgement.
• Scopes in the PR body match folders in the diff.
• Primary-doc links were smoke-checked after publishing edits.
• MCP connectors mentioned (if any) list owners.

Boundary Note

If your repo cannot state boundaries plainly, agents will guess—and guessing scales poorly.

Source Stack

• Claude Code — getting started
• Claude Code — skills
• Claude Code — hooks
• Claude Code — permissions
• Model Context Protocol — specification
• Anthropic — Claude Code product page
• Anthropic Claude Code releases

Next Move

Take the next step on Claude Code team conventions—walk the exercises once with receipts enforced, not once with vibes enforced.