The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python utility scripts/atoms-search.py using python3 to search its internal knowledge base.
[EXTERNAL_DOWNLOADS]: The search script connects to https://api.xyskill.xyz/v1/atoms/search to fetch the latest knowledge entries. This is a vendor-owned resource managed by the author 'xyaz1313'.
[DATA_EXFILTRATION]: The atoms-search.py script sends the user-provided search query to the remote API. If the query contains sensitive information provided in the agent's context, that data is transmitted to the external server.
[DATA_EXPOSURE]: The script scans the user's home directory for configuration and data files, specifically checking ~/.xy/config.json and various skill installation paths like ~/.claude/skills/xy to locate its database.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data such as video transcripts and product descriptions.
Ingestion points: User-provided video scripts, links, and product details in Phase 1.
Boundary markers: The skill implements an 'evidence grading' system (信条 2) and a 'selling point gate' (信条 4) to ensure the agent distinguishes between facts and claims.
Capability inventory: The skill has the ability to execute shell commands and perform network operations via the atoms-search.py script.
Sanitization: There is no evidence of explicit sanitization or escaping of the user-provided content before it is processed by the agent or passed to the search script.

xy-replica