The Agent Skills Directory

[SAFE]: The skill correctly implements its documented purpose of writing JSONL data to Amazon S3 using the official Boto3 library.- [SAFE]: Input validation is performed on S3 bucket names and prefixes using regular expressions in src/sink.py to prevent malformed requests or path injection.- [SAFE]: The execution flow is safe by default, using a dry-run mode unless the --apply flag is explicitly provided by the user.- [SAFE]: The skill uses standard AWS SDK credential management patterns, ensuring that secrets are managed via environment variables or workload identity rather than being hardcoded.- [PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data that enters via sys.stdin. \u2022 Ingestion points: Data is read from sys.stdin in src/sink.py and converted to JSON. \u2022 Boundary markers: The data is treated as an opaque payload and is not delimited from instructions. \u2022 Capability inventory: The skill has the capability to write objects to S3 using the put_object API. \u2022 Sanitization: The script validates that input is well-formed JSON using json.loads before processing, ensuring structural integrity of the sink operation.

sink-s3-jsonl