The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md and the wrapper script scripts/run_osint.sh direct the agent to execute shell commands where user-provided input is interpolated directly into the command string (e.g., python3 -m social-analyzer --username "{USERNAME}"). This creates a risk of command injection if the input contains shell metacharacters that are not properly escaped.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from 1000+ external social media platforms, including user biographies and display names, which are then processed by the agent to generate a summary.
Ingestion points: External profile metadata retrieved by social-analyzer and scripts/cn_lookup.py from platforms like Bilibili, Zhihu, and Weibo.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the retrieved profile content as untrusted data, increasing the risk that instructions embedded in a profile bio could influence agent behavior during summarization.
Capability inventory: The skill is configured to use python3 and pip3 for local execution and package management, providing a functional environment that could be targeted if an injection succeeds.
Sanitization: The profile data is presented to the agent for natural language processing without prior sanitization or validation of the content.

osint-social