The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input via the --question parameter, which accepts the customer's original wording. This data is used by scripts/_doc_lookup.py to match against a documentation index for verification purposes.
Ingestion points: The question parameter in scripts/multipart_upload_diagnosis.py which is passed to the lookup_config_topic function in scripts/_doc_lookup.py.
Boundary markers: The skill instructions provide strict "Absolute Rules" prohibiting the execution of mutating OSS APIs and requiring the use of provided scripts for all diagnostic activities.
Capability inventory: The skill utilizes the official oss2 Python SDK for metadata queries and the aliyun CLI for identity verification via STS.
Sanitization: Input is primarily used for keyword matching within the documentation lookup script; shell commands are executed using structured argument lists via subprocess.run to prevent shell injection attacks.
[EXTERNAL_DOWNLOADS]: The skill fetches a documentation index (llms.txt) and markdown excerpts from help.aliyun.com. This is an official Alibaba Cloud service used to provide verified information and citations in response to customer queries.
[COMMAND_EXECUTION]: The skill invokes the official aliyun CLI tool (specifically the aliyun sts get-caller-identity command) to verify the caller's identity. It also includes documentation suggesting the agent upgrade the CLI using aliyun upgrade. These operations are performed using safe subprocess calls with argument lists.

alibabacloud-oss-multipart-upload-diagnosis