# Issue: migrate Improv Rating Wagers
Handle: `improv-rating-wagers-frontier-cs-algorithmic-77`
Source: `algorithmic/problems/77`
## Provenance
- Frontier-CS category: `algorithmic`
- Source path: `algorithmic/problems/77`
- Source files inspected: `config.yaml`, `statement.txt`, `interactor.*`, and `testdata/`
- Original execution shape: interactive-style Frontier-CS benchmark
## Agentics Migration
- Mode: `separated_evaluator`
- Target: `linux-arm64-cpu`
- Solution protocol: `zip_project` with stdin/stdout
- Evaluator: Python exact-reference checker that writes Agentics `result.json`
- Public validation: one tiny committed smoke run with answer `0`
- Private benchmark: `3` official Frontier-CS-derived run(s) in `official-runs.zip`
## Public/Private Split
Public bundle contains metadata, statement, evaluator code, and tiny validation data only. Official source-derived run records and reference answers are stored in the private overlay at `private-benchmark/runs.json` plus private answer files.
## Acceptance Criteria
- `python3 scripts/validate_challenges.py` passes in the challenge worktree.
- Public smoke evaluator completes and records a valid Agentics result JSON.
- Private overlay ZIP contains no symlinks or unsafe traversal paths.
- No official private cases or reference answers are committed to Git.
## Known Risks
- The original task was interactive; this migration intentionally uses an offline exact-reference stdin/stdout contract, so query-efficiency behavior is not preserved.