LIVE-13 — Recording the Grok Capability Retest Accurately

The broader goal at the end of the previous PR

PR #2035 corrected Grok's stale gh-write capability policy after the live TC-9 retest passed. It also changed the non-live probe path so an unmeasured result would no longer be reported as a sandbox denial. The remaining incident work was to close the corresponding action in the LIVE-13 RCA with an accurate record.

Strategic shifts in this PR

There was no product or architecture shift. The documentation needed to record what the evidence actually established. The earlier gh pr list succeeded through Grok's headless authenticated CLI path, proving read access. Separately, job-20114270 created issue comment #5986713400 at 2026-10-05T01:46:34Z through the authenticated headless gh CLI. That is direct write evidence, distinct from the earlier read, and the RCA now records both observations without conflating them.

What this PR ships

The LIVE-13 RCA marks its TC-9 capability retest complete, links the correction to PR #2035 and gh:#2034, and records the successful sandbox_allowed result, the earlier read, and the separate literal issue-comment write. No runtime code or data format changed.

What was achieved toward the long-arc goal

The capability record now distinguishes a live authenticated GitHub CLI success from a literal write sample. That keeps the incident history useful for later routing decisions while keeping the one direct write sample distinct from the read result.

New goalpost

Grok remains last in the gh-write priority tuple. This single successful issue comment does not change the empirical policy's sample-size requirements or promote routing; it records only the newly verified direct-write evidence.

No deploy or migration required.