ci: Redline caller pinned to askalf/askalf@116935d (#64, tool names the gateway forwards verbatim)

[askalf]

What

.github/workflows/redline.yml calls askalf/askalf/.github/workflows/redline-review.yml at a fixed commit. This moves the pin from 35a0ee3 to 116935d, which carries askalf/askalf#64: the four review tools are now redline_list, redline_read, redline_search and redline_submit.

Why

With the old names, dario's tool map recognised grep as its own Grep and dropped the other three on the Claude path for some seats, so the model answered "the only tool available is Grep" and could never submit. Two of the four runs at 35a0ee3 on 2026-09-25 20:1xZ died exactly that way (#65, amnesia#92) while the pool was healthy. Names the gateway cannot map make it treat the client as non-CC and forward the tools verbatim.

Nothing else changes

Same runner label, same guard, same permissions. The workflow is a pin bump only.

[sprayberry-redline]

Automated review from the Sprayberry Labs fleet code reviewer.

Reviewed by the gating lane (gating review).

Verdict: approve. Approved. The only change is the uses: ref for the shared review workflow moving from 35a0ee3805b3… to 116935d3803f…, both full 40-character commit SHAs, so the pin stays immutable. I checked the rest of the caller workflow for regressions: the trigger is still pull_request (not pull_request_target), permissions remain contents: read / pull-requests: read, the same-repo and non-draft guard on the job is unchanged, and the only input is the runner label. The PR body and commit message are plain prose with no AI attribution, secrets, or internal hostnames. I cannot inspect the callee at the new commit from this checkout, but a SHA-pinned bump through a reviewed PR is exactly the mechanism the comment on line 22 describes.