The Agent Skills Directory

[COMMAND_EXECUTION]: The sync scripts sync_telegram_export.py and sync_telegram_saved.py use subprocess.run to invoke the companion script archive_x.py. The implementation uses list-based arguments which prevents shell command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from X (via fxtwitter) and Telegram, which are untrusted external sources. The AI is instructed to generate summaries and translations of this content.
Ingestion points: Data is ingested via archive_x.py fetching JSON from api.fxtwitter.com and reading Telegram export/API data.
Boundary markers: The skill uses Markdown headings (e.g., ## 摘要, ## 原文) to separate AI-generated content from source text, but lacks explicit instructions to ignore embedded commands within the source data.
Capability inventory: The agent has capabilities to write to the local file system (Obsidian vault) and execute local scripts via subprocess.run.
Sanitization: There is no evidence of sanitizing the scraped content before the AI is asked to process it for summarization or translation.
[EXTERNAL_DOWNLOADS]: The skill documentation and scripts (generate_telegram_session.py) recommend the installation of the telethon Python package to enable Telegram MTProto synchronization features.

soia-pkm-clip-x