The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions define specific constraints and scoping rules to guide the agent's behavior for mathematical analysis. There are no attempts to bypass safety filters or override system instructions for malicious purposes.
[DATA_EXFILTRATION]: No network operations or sensitive file paths were identified. The skill does not instruct the agent to send data to external servers or access private credentials.
[REMOTE_CODE_EXECUTION]: The skill does not contain any commands to download or execute external scripts, binaries, or packages.
[COMMAND_EXECUTION]: There are no shell commands, subprocess calls, or system-level interactions described in the instructions.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process untrusted data (research papers), it lacks any dangerous capabilities (e.g., file writing, network access, or command execution) that could be exploited by an injection attack embedded within a paper. The scope of the agent's actions is limited to generating text-based summaries and analysis.
[OBFUSCATION]: The content is written in clear, standard markdown. No hidden characters, encoded strings, or obfuscated URLs were detected.

gap-finder