ci(seidroid): bump uci to v0.0.26

[bdchatham]

Summary

Bumps the seidroid workflows to uci v0.0.26.

• Reviews run on sei-agent-driver v0.23.0 at high reasoning effort.
• A re-run after a no-verdict starts in a fresh session, instead of repeating the refused reply.
• The AI Assistant no longer answers @seidroid review close with its own review.

Change

• seidroid.yml: the three jobs move from v0.0.25 (fcdcbed) to v0.0.26.
• ai-assist.yml: uses: and uci-ref move from v0.0.19 (eb65d6d) to v0.0.26. Between those versions ai-assistant.yml changed only by fix(ai-assistant): reserve @seidroid review close for the review workflow uci#117, and its inputs are unchanged.

Every input this repository passes still exists in v0.0.26. The only input change is the new claude-effort, which defaults to high.

Verification

actionlint on the changed files: clean.

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Workflow pin updates only; permissions, secrets, and passed inputs are unchanged.

Overview
Pins reusable sei-protocol/uci workflows to v0.0.26 (0dc48cfb) in CI wiring only—no new inputs are passed from this repo.

seidroid.yml: All three jobs (seidroid-review, seidroid-review-close, seidroid-review-reclaim) move from v0.0.25 to v0.0.26, picking up upstream behavior such as sei-agent-driver v0.23.0 at high reasoning effort, fresh sessions on re-run after a no-verdict, and related review fixes.

ai-assist.yml: The uses: ref and uci-ref input jump from v0.0.19 to the same v0.0.26 commit so the AI Assistant stays aligned with seidroid; upstream change is mainly that @seidroid review close is no longer answered with a spurious review.

^{Reviewed by Cursor Bugbot for commit a0e4cdd. Bugbot is set up for automated code reviews on this repo. Configure here.}

[seidroid]

This PR moves all four uci reusable-workflow pins (three seidroid-review jobs and the AI Assistant's uses: plus uci-ref) to 0dc48cf, the commit the v0.0.26 tag points to. I checked the pins against the tag, against the inputs each caller passes and against the callee job permissions, and found no problems; the codex reading also reported nothing, so there was nothing from it to keep or drop.

_{seidroid review · decision approve · session a98004c239e34e9089e8556b8b0b67a2 · turn resp_claude_b43b3698edad455207f293c9de160692 · item fa50eac90a855e678d3f0415086ced23}

Findings: 0 blocking | 0 non-blocking | 0 posted inline