The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes WebFetch and a reachability script to connect to official TÜBİTAK (Scientific and Technological Research Council of Turkey) and ULAKBİM (National Academic Network and Information Center) domains. These represent well-known and trusted academic services appropriate for the skill's intended purpose.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute two local Python scripts, vyp_scaffold.py and policy_check.py. These scripts are used for generating Markdown templates and checking URL status, respectively. They rely on standard library features and include optional support for the requests package, with no evidence of dangerous system commands or unauthorized process spawning.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes information from external scientific policy websites, establishing a potential attack surface. However, this is addressed by the following evidence chain:\n
Ingestion points: External policy URLs are fetched via WebFetch and the policy_check.py script.\n
Boundary markers: The SKILL.md instructions include a mandatory 'Verify-Against-Current-Policy Disclaimer' that prompts the user to verify all output against live official PDF sources.\n
Capability inventory: The skill has access to Write, Edit, and Bash for managing project files and running local helpers.\n
Sanitization: External data is processed solely for metadata validation and keyword checking (e.g., checking for the 'aperta' string) rather than for dynamic code assembly or complex parsing.

alterlab-aperta