The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The atoms-search.py script attempts to fetch data from https://api.xyskill.xyz/v1/atoms/search. This domain is associated with the skill author's infrastructure and is used for knowledge retrieval.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python utility via a shell command: python3 <path>/atoms-search.py "<关键词>". This creates a potential risk of command injection if the <关键词> placeholder is filled with unsanitized user input, which could allow for the execution of arbitrary shell commands in the agent's environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (wishes and goals) and incorporates results from an external search script into its prompt context.
Ingestion points: User dialogue inputs described in the SKILL.md workflows.
Boundary markers: The skill uses structured Markdown tables and specific "Goal Card" templates to organize extracted data, which provides some structural separation.
Capability inventory: The skill uses shell execution of Python scripts, local file system reads of atoms.jsonl, and network requests to api.xyskill.xyz.
Sanitization: There are no explicit instructions or logic within the skill to sanitize or escape content retrieved by the atoms-search.py script before it is processed by the agent.
[DYNAMIC_EXECUTION]: The core functionality of the skill relies on the runtime execution of the atoms-search.py script to perform dynamic knowledge lookups, which constitutes dynamic code invocation at runtime.

xy-goal-card