Prompt
Workspace files:
project/backups/snapshot.index.json
project/backups/snapshots/base_20260310/manifest.json
project/backups/snapshots/inc_20260311/manifest.json
project/backups/snapshots/inc_20260312/manifest.json
project/scripts/chain_probe.py
The most recent incremental backup verification has failed. Your tasks:
- Start by inspecting the index, manifests, and payload files to determine the true root cause — do not just look at the last error message.
- Fix only the necessary backup metadata so the latest snapshot chain is restorable again.
- Do not modify any payload data files.
- Do not modify
chain_probe.py.
- Run
python project/scripts/chain_probe.py from the workspace root and make sure it outputs CHAIN_OK and writes output/backup_restore_probe.json.
- Write a report at
BACKUP_REPAIR.md (in the workspace root) explaining:
- The true root cause.
- Which errors were merely symptoms.
- What index/metadata changes you made.
- How you verified the fix.
Constraints:
- Limit your changes to
project/backups/snapshot.index.json as much as possible.
- Do not modify any log files (if any exist).
- Do not fabricate recovery results.
Expected Behavior
- Inspect
snapshot.index.json and observe that inc_20260312.parent is incorrectly set to base_20260310.
- Restoring through that parent chain skips
inc_20260311 and therefore loses:
app/config.yml v2 revision (only present in inc_20260311/payloads/app_config_v2.yml).
reports/daily_20260311.csv (only present in inc_20260311).
- Fix: change
inc_20260312.parent from base_20260310 to inc_20260311.
- Run
python project/scripts/chain_probe.py and verify it prints CHAIN_OK and writes output/backup_restore_probe.json with:
{
"latest_snapshot_id": "inc_20260312",
"restored_files": [
"app/config.yml",
"data/customers.csv",
"exports/latest_summary.txt",
"reports/daily_20260311.csv",
"reports/daily_20260312.csv"
],
"config_revision": "v2",
"status": "healthy"
}
- Author
BACKUP_REPAIR.md with the root cause, symptoms, fix, and verification.
Grading Criteria
Workspace Files
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshot.index.json -> project/backups/snapshot.index.json
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/base_20260310/manifest.json -> project/backups/snapshots/base_20260310/manifest.json
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/base_20260310/payloads/app_config_v1.yml -> project/backups/snapshots/base_20260310/payloads/app_config_v1.yml
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/base_20260310/payloads/customers_base.csv -> project/backups/snapshots/base_20260310/payloads/customers_base.csv
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/inc_20260311/manifest.json -> project/backups/snapshots/inc_20260311/manifest.json
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/inc_20260311/payloads/app_config_v2.yml -> project/backups/snapshots/inc_20260311/payloads/app_config_v2.yml
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/inc_20260311/payloads/daily_20260311.csv -> project/backups/snapshots/inc_20260311/payloads/daily_20260311.csv
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/inc_20260312/manifest.json -> project/backups/snapshots/inc_20260312/manifest.json
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/inc_20260312/payloads/daily_20260312.csv -> project/backups/snapshots/inc_20260312/payloads/daily_20260312.csv
assets/T051_claweval_CTB_W05_backup_chain_repair/project/backups/snapshots/inc_20260312/payloads/latest_summary.txt -> project/backups/snapshots/inc_20260312/payloads/latest_summary.txt
assets/T051_claweval_CTB_W05_backup_chain_repair/project/scripts/chain_probe.py -> project/scripts/chain_probe.py
Platform Delivery
This is the Jingxuan Arena single-task adaptation of an agentscope-ai/PawBench benchmark task. Produce the required workspace files, summaries, or structured outputs exactly as the prompt requests. Official scoring is computed by the platform, and the public task page intentionally omits raw automated checks, hidden judge rubrics, and reference answers.
Task Metadata
- Source:
PawBench v1.0
- Source Dataset:
ClawEval
- Source Task ID:
CTB_W05_backup_chain_repair
- Grading Type:
Hybrid
- Timeout:
600 seconds
- Scenario:
Software Engineering Devops
- Capabilities:
Logic Reasoning, Tool Use, Planning, Self Verification
- Complexity:
L3
- Environment:
Open
- Modality:
Text