docs: forbid customer/private identifiers in this public repo

[steventohme]

Why

This repo is public. Customer/company names and other private identifiers have leaked into router PRs before (e.g. an org name in a loop-fix PR title/description). We want a standing rule that agents and humans see up front.

What

Adds one bullet to the top of the "Things to NEVER do" section, mirrored verbatim in both AGENTS.md and CLAUDE.md (per the files' mirror notice):

Never put customer/company/org names or private identifiers in anything committed here. This repo is public — commit messages, branch names, PR titles/descriptions, code comments, tests, and fixtures are world-readable. This holds even when a change is motivated by one customer's investigation … describe the trigger generically ("a customer", "a large agentic session", "an org's monorepo"). No org names, org IDs, account emails, internal ticket/Linear links, or verbatim private Slack/support excerpts — those stay in the private WorkWeave repo + Linear.

Instruction-layer guardrail only — it loads into any agent working in a router checkout. A mechanical backstop (pre-push/CI grep against a private denylist) can follow if we want a hard block.

🤖 Generated with Claude Code

[greptile-apps]

T-Rex Logs

What T-Rex did

• The initial validation attempt was captured in docs_guardrail-01-before.log, recording placement and byte-for-byte mirroring and showing that the exact phrase check was not satisfied by the actual wording.
• The corrected validation attempt is captured in docs_guardrail-02-after.log, showing exit code 0 along with line numbers, extracted bullet text, immediate-placement result, and the complete byte-for-byte mirror and guardrail content checks being satisfied.

_{Ran code and verified through T-Rex}

_{Reviews (1): Last reviewed commit: "docs: forbid customer/private identifier..." | Re-trigger Greptile}