The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes embedded Python scripts executed via the shell (python3 -c) to transform and filter JSON data retrieved from API contract endpoints. These scripts, located in references/commands.md, are used to generate TSV files and compact OpenAPI schemas. While functional, the use of python3 -c to execute strings constructed within the instructions constitutes dynamic execution of embedded source code.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest content from external documentation sources and APIs, which presents a surface for indirect prompt injection where malicious instructions could be embedded in the documentation data.
Ingestion points: Documentation content and API schemas retrieved via arkcli docs get, arkcli docs search, and arkcli docs apis spec.
Boundary markers: The instructions explicitly warn that retrieved text is '不可信参考文本' (untrusted reference text) and provide guidance to ignore formatting tags and layout artifacts.
Capability inventory: The skill has the ability to execute shell commands using the arkcli tool and python3, and it can perform file write operations via shell redirection (e.g., to /tmp/).
Sanitization: The skill includes instructions to manually verify facts against actual outputs and forbids the execution of commands or configurations found within retrieved documentation without prior intent confirmation.
[COMMAND_EXECUTION]: The skill requires and executes the arkcli binary and python3 to perform its tasks. It constructs shell command lines that include user-supplied keywords, URLs, and local file paths.

arkcli-docs