The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documentation and the logic implemented in scripts/query_rankings.py automate the retrieval of secrets from sensitive system configuration files. It specifically encourages users to store API keys in shell profiles and then programmatically scans those files to extract the keys.
[DATA_EXFILTRATION]: scripts/query_rankings.py performs unauthorized reads of sensitive file paths, including ~/.bashrc, ~/.zshrc, ~/.bash_profile, ~/.profile, and Windows PowerShell profiles, to harvest environment variable values. Additionally, scripts/export_to_excel.py and scripts/generate_ranking_image.py interact with the user's desktop directory (~/Desktop) to save and copy files.
[EXTERNAL_DOWNLOADS]: The skill makes network connections to the vendor-owned domain redfox.hk and the well-known service api.weixin.qq.com for data retrieval and notifications.
[INDIRECT_PROMPT_INJECTION]: The skill presents a vulnerability surface by ingesting data from an external API (redfox.hk) and providing capabilities to write files to the user's desktop. The ingestion process in scripts/query_rankings.py is insecure as it explicitly disables SSL hostname checking and certificate verification (ssl.CERT_NONE), making the system vulnerable to Man-in-the-Middle (MITM) attacks where malicious data could be injected.
Ingestion points: API response processed in scripts/query_rankings.py.
Boundary markers: None detected.
Capability inventory: File write and copy operations to ~/Desktop in scripts/export_to_excel.py and scripts/generate_ranking_image.py.
Sanitization: No sanitization or validation of the ingested external content was observed before file operations.

xhs-follower-growth-chart