The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The instructions in SKILL.md and references/data_sources.md direct the agent to execute npx -y socialdatax-skills@latest. This command downloads a package from the NPM registry and executes it immediately, presenting a significant risk of executing unverified code at runtime.
[COMMAND_EXECUTION]: The skill relies on shell command execution to perform its tasks. It executes local Python scripts (scripts/fetch_xhs_hot_articles.py, scripts/compare_sets.py) and Node.js tools (node $XT/src/xiaohongshu/search-cli.js).
[EXTERNAL_DOWNLOADS]: The skill makes network requests to external, non-trusted data providers including redfox.hk, socialdatax.com, and guaikei.com to retrieve Xiaohongshu data.
[CREDENTIALS_UNSAFE]: The skill instructions ask the user to configure sensitive API keys (REDFOX_API_KEY, SOCIALDATAX_API_KEY, GUAIKEI_API_TOKEN) in environment variables, which are then accessed by the agent and potentially passed to external CLI tools.
[DATA_EXFILTRATION]: User-provided search keywords and query parameters are sent to external third-party API endpoints during the retrieval process.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it ingests and processes content from Xiaohongshu (titles, descriptions, comments) while retaining the capability to execute shell commands.
Ingestion points: Note titles, descriptions, and comments retrieved via API responses from XHS.
Boundary markers: No explicit boundary markers or directives are used to separate external data from agent instructions.
Capability inventory: The agent has shell execution capabilities including python3, node, and npx.
Sanitization: No sanitization or validation of the retrieved external content is performed before processing.

space-xhs-hotspot