test: add knowledge-bot evolution cases

[upsearchmain]

Summary

• Add cases/knowledge-bot/cases.jsonl with 30 benchmark cases.
• Distribution: book-deep 10, rule-extract 10, decision-30day-review 10.
• Keep a consistent JSONL shape with id, skill, category, input, expected, and judge.

Verification

test -f ~/hermes-evolution/cases/knowledge-bot/cases.jsonl && echo exists
# exists

wc -l ~/hermes-evolution/cases/knowledge-bot/cases.jsonl
# 30 /home/claude/hermes-evolution/cases/knowledge-bot/cases.jsonl

/usr/bin/python3 -c "import json, pathlib; [json.loads(l) for l in pathlib.Path('$HOME/hermes-evolution/cases/knowledge-bot/cases.jsonl').read_text().splitlines()]" && echo OK
# OK

git diff --check origin/main...HEAD
# exit 0

Additional distribution check:

/usr/bin/python3 -c "import collections,json,pathlib; rows=[json.loads(l) for l in pathlib.Path('$HOME/hermes-evolution/cases/knowledge-bot/cases.jsonl').read_text().splitlines()]; print(len(rows)); print(dict(sorted(collections.Counter(r['category'] for r in rows).items()))); assert len(rows)>=30; assert collections.Counter(r['category'] for r in rows)=={'book-deep':10,'rule-extract':10,'decision-30day-review':10}; assert all(set(['id','input','expected','judge']).issubset(r) for r in rows)"
# 30
# {'book-deep': 10, 'decision-30day-review': 10, 'rule-extract': 10}

Self-Review

• P1 Think Before Coding: Read AGENTS, task, v5 PART 11/PART 13, book-deep, edge-cases, and acceptance section 7 before editing.
• P2 Simplicity First: Data-only JSONL change; no generator, package install, LLM/API calls, or extra schema files.
• P3 Surgical Changes: Only cases/knowledge-bot/cases.jsonl is added in this repo.
• P4 Goal-Driven Execution: Existence, line count, JSON parse, whitespace, shape, and category distribution were verified.