Download Results Skill

Purpose

⚠️ Configuration Protection Policy

CRITICAL: Never modify ~/.hotvect/ configuration files. See CONFIG_PROTECTION_POLICY.md for full policy.

This skill reads ~/.hotvect/config.json for configuration but never modifies it. Work only in user-specified directories.

Automatically download SageMaker backtest results from S3 when user needs to analyze training outputs or compare algorithm performance.

Configuration

CRITICAL: You (Claude) must read ~/.hotvect/config.json at skill invocation to construct complete hv-ext download-results commands.

How this works:

• •The hv-ext download-results CLI tool does NOT read config.json
• •YOU read config.json and construct the full command line with all arguments
• •Config provides defaults that can be overridden by user-specified values

When this skill activates:

1. •

   Read config: cat ~/.hotvect/config.json

2. •

   Parse the JSON to extract values (you can understand JSON directly)

3. •

   Use config values for AWS operations:

   • •aws.credential_helper → Command to run when refreshing credentials
   • •directories.output_base_dir → Default for --dest-base-dir argument
4. •

   Fail fast if config doesn't exist and AWS operations are needed

5. •

   Tell user to run /agent hotvect-setup-agent if config is missing

Important: Always construct complete hv-ext download-results commands with ALL required arguments explicitly specified.

When to Invoke

Invoke this skill when:

• •User mentions downloading backtest results
• •User asks to check SageMaker outputs
• •User wants to analyze training results from S3
• •User completed a backtest and needs the outputs

Context Requirements Before Activation

Before activating this skill, ensure you have gathered:

1. •Date range: From-date and to-date for results (ask user if not mentioned)
2. •S3 base prefix: Standard location or user-specified
3. •Destination directory: From config directories.output_base_dir or user-specified
4. •AWS credentials status: Check if credentials are valid (run aws sts get-caller-identity)

If critical information is missing:

• •Ask user for date range before proceeding
• •Don't activate skill until you have enough information to construct a complete hv-ext download-results command

Operations

1. Identify Download Requirements

From user's message, determine:

• •Date range for results (from-date, to-date)
• •Specific backtest run (if mentioned)
• •Destination directory (default: ./backtest-results)

2. Check AWS Credentials

Verify AWS access:

aws sts get-caller-identity

If expired or missing, guide user to refresh:

zalando-aws-cli login my-team-data ReadOnly

3. Determine S3 Location

Standard location for SageMaker backtest results:

s3://my-experiment-bucket/temp/username/sagemaker_output/

Or user-specific path if mentioned.

4. Construct Download Command

Build hv-ext download-results command:

hv-ext download-results \
  --s3-base-prefix s3://my-experiment-bucket/temp/username/sagemaker_output/ \
  --dest-base-dir ./backtest-results \
  --from-date ${start_date} \
  --to-date ${end_date} \
  --include-metadata

5. Execute Download

Run the command and monitor progress:

• •Show dates being downloaded
• •Report file counts and sizes
• •Confirm successful completion

6. Verify Downloaded Results

After download, check:

ls -lh ./backtest-results
tree -L 2 ./backtest-results

Report:

• •Number of dates downloaded
• •Files per date
• •Total size
• •Location of result files

7. Suggest Next Steps

After successful download, suggest:

• •Compare evaluation results: hv-ext compare-evaluations
• •Review specific outputs
• •Analyze performance metrics

Expected Output

Successful download:

Downloading backtest results from S3...

Date range: 2025-06-01 to 2025-06-15
Destination: ./backtest-results

Downloaded:
- 2025-06-01: 3 files (1.2 GB)
- 2025-06-02: 3 files (1.1 GB)
...
- 2025-06-15: 3 files (1.2 GB)

Total: 45 files, 17.8 GB

Results available in: ./backtest-results

Error Handling

AWS credentials expired:

• •Run zalando-aws-cli login my-team-data ReadOnly
• •Retry download

S3 path not found:

• •Verify S3 base prefix
• •Check if backtest completed successfully
• •Confirm date range has data

Network issues:

• •Retry with --max-attempts flag
• •Download in smaller date ranges

Insufficient disk space:

• •Report space required
• •Ask user to free space or change destination

Communication

• •Show download progress clearly
• •Report estimated total size before starting
• •Confirm successful completion
• •Suggest next analysis steps