The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/create_function.py accepts a file path via the --code argument and reads its entire content using f.read(). This content is then encoded and sent to Huawei Cloud via an API request. This mechanism, while intended for uploading function code, creates a risk where a user or an injected prompt could trick the agent into reading and exfiltrating sensitive local files (e.g., ~/.ssh/id_rsa, .env files) by providing their paths as the code source.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could contain malicious instructions designed to influence the agent or the deployment environment.
Ingestion points: scripts/create_function.py reads local files via --code and parses JSON payloads from --env-vars and --body arguments.
Boundary markers: No specific boundary markers or 'ignore' instructions are used to isolate untrusted data from the agent's logic.
Capability inventory: The skill possesses the ability to read arbitrary files from the local filesystem and perform network operations via the Huawei Cloud SDK.
Sanitization: The script performs basic validation on metadata (e.g., verifying function names are alphanumeric and memory sizes are within range) but lacks any sanitization or inspection of the actual code content or JSON payload values.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute Python scripts (scripts/create_function.py) with various command-line arguments. While these scripts are part of the skill, the execution of shell commands based on user-influenced parameters is a standard surface for potential abuse.

huawei-cloud-functiongraph-function-create