The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted data from sucai.md and external results from the web_search tool to populate article content. This creates a surface for indirect prompt injection, as the skill lacks explicit boundary markers or instructions to the agent to disregard potential instructions embedded in the source material.\n
Ingestion points: sucai.md (user-provided materials) and results from web_search.\n
Boundary markers: No specific delimiters or "ignore previous instructions" warnings are implemented for processed data.\n
Capability inventory: The skill has access to web_search, bash (executing the date command and a local Python script), and AskUserQuestion.\n
Sanitization: There is no evidence of filtering or sanitization of content retrieved from external sources.\n- [COMMAND_EXECUTION]: The skill uses the bash environment to execute the date command for timestamping articles and runs a bundled script ./scripts/check_body_length.py to verify the word count of the generated article. These commands are used for utility purposes and are restricted to specific, non-malicious tasks within the local execution environment.

job-article