task-observer: new docs/ai/experiment-diagnostics from 2026-07-01

[divad12]

Target\nnew-global-doc: docs/ai/experiment-diagnostics.md\n\n## Observations addressed\n- Observation 13 (journology/log.md): Probe loops should promote implausible guidance to bugs — changed: created docs/ai/experiment-diagnostics.md establishing the contract that when a probe produces implausible user-facing guidance, the experiment queue pauses, the diagnostic source is fixed (with a behavior test), the canonical probe is rerun, and only corrected artifacts are promoted to durable evidence. Also added a routing table row to .claude/AGENTS.md so agents load this doc when running probe loops, experiment diagnostics, or tuning sessions.

---

Generated by Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7453fa0c88

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Add the docs/ai load marker

When probe/tuning tasks hit the new routing row, this file is meant to be a canonical adaptive-docs reference, but it starts directly with the heading instead of the standard Context loaded: experiment-diagnostics.md probe. docs/ai/writing-docs.md says that top-of-file probe is how the harness verifies which docs/ai files actually loaded, so this new contract loses the observability that lets users and agents confirm the right guidance was read before experiment artifacts are promoted.

Useful? React with 👍 / 👎.