fix(live-proof): install the target toolchain and assert satisfiable values

[steipete]

Summary

Two findings from the first real runs of the merged one-job architecture.

Toolchain. openclaw/clawhub#3457 failed before its first step with sh -lc bun install --ignore-scripts failed: sh: 1: bun: not found — the review runner ships pnpm, and ClawHub is a bun repository. The target's package manager is now installed on demand from the repository profile's package_manager, only after the verification gates pass so ordinary reviews stay lean, and an installer failure publishes a failed verification result rather than failing the review.

Assertion quality. Both first merged runs asserted values their demonstrations could never satisfy: #1182 filled the Bay finder with an invented 97722 and asserted Found, and clawhub asserted the literal 52 passed from a test suite. The prompt now requires:

Every assertion must name something the demonstration can actually satisfy. Assert values that the page or command will genuinely produce: for a search box, search for a value the page itself already displays; for a command, assert a stable substring of its output such as a header, flag name, or error string, not a count, timing, or number that varies per run.

Worth noting what #1182 already proved despite that failure: press / and wait_for #finder-input:focus both PASSED, verifying live that the keyboard shortcut the PR adds actually works.

Validation

• pnpm build clean; live-proof 52/52; prompt-policy 39/39; full unit suite 2,347 passed / 0 failed.
• Bun install command composed as curl -fsSL https://bun.sh/install | bash.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

[clawsweeper]

ClawSweeper status: review started.

I am starting a fresh review of this pull request: fix(live-proof): install the target toolchain and assert satisfiable values This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.