The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze public TikTok data, which constitutes an untrusted data source.
Ingestion points: External data from TikTok profiles, video captions, and comments (fetched via api.crawlora.net) are provided to the agent as part of the research workflows described in SKILL.md.
Boundary markers: The instructions lack explicit boundary markers or instructions to treat the ingested TikTok content as data rather than instructions.
Capability inventory: The skill has the capability to execute network-enabled shell scripts (scripts/crawlora.sh).
Sanitization: No sanitization or filtering logic is present to handle potentially malicious payloads embedded in TikTok comments or bios.
[COMMAND_EXECUTION]: The skill relies on a local bash helper script to interact with external APIs.
Evidence: scripts/crawlora.sh is invoked to perform GET requests to the Crawlora API.
Context: The script demonstrates high security awareness by using umask and chmod to protect API keys in temporary files, validating paths against a strict allow-list, and using curl's URL-encoding features to prevent argument injection.

tiktok-research