The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell command execution to interact with the host system's credential manager.
Evidence: The script scripts/tikhub_request.py invokes the security utility on macOS to retrieve API keys from the Keychain. The command is executed using a structured list of arguments without a shell, minimizing injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted social media content (posts, comments) which acts as a potential surface for indirect prompt injection.
Ingestion points: Social media data is ingested from TikHub API responses and stored in the social-research/raw/ directory via scripts/tikhub_request.py.
Boundary markers: Instructions in SKILL.md and references/output-schema.md specify clear separation between raw evidence and AI-generated analysis fields.
Capability inventory: The skill can make network requests to the TikHub API host and write results to the local file system.
Sanitization: The tikhub_request.py script includes a dedicated redact_payload function that recursively filters sensitive keys like api_key, token, and authorization from all displayed outputs and logs.

shengjiang-research