Add fix agent human /fs-fix functional eval

[ascerra]

Summary

Add a functional eval under eval/fix/ that runs the real fix agent pipeline against an existing PR fixture (human /fs-fix via env) and asserts the post-script pushes a new commit touching calc.py.

Motivation

Regression guard for the fix pipeline (human trigger path). Code eval (#180 / PR #177) covers issue→PR; fix needs existing-PR→new-commit.

Acceptance criteria

• eval/fix/ with at least one human /fs-fix case
• Shared eval scripts supply fix env (HUMAN_INSTRUCTION, PRE_AGENT_HEAD, empty REVIEW_BODY_FILE) and checkout PR head
• Capture includes head_sha, files, pre_agent_head for judges
• Functional-tests CI selects and runs fix

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 3:17 PM UTC · Completed 3:19 PM UTC
Commit: 5cd495a · View workflow run →

[fullsend-ai-triage]

Triage Summary

PR #183 is already open and directly addresses this issue (includes Closes #182). It adds:

• eval/fix/ with the 001-human-fs-fix-add case
• Shared eval script extensions for fix env (HUMAN_INSTRUCTION, PRE_AGENT_HEAD, empty REVIEW_BODY_FILE) and PR-head checkout
• Capture of head_sha, files, and pre_agent_head for judges
• new_commit and expected_files judges in eval/fix/eval.yaml

This covers all four acceptance criteria. PR #183 also notes a dependency on PR #177 (code agent functional eval) for shared script changes.

No additional implementation work should be dispatched until PR #183 is resolved.

Blocked by:

• test(eval): add fix agent human /fs-fix functional eval #183