The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file contains an instruction to execute a shell command for generating diagrams: python scripts/generate_schematic.py "your diagram description" -o figures/output.png. This pattern directly interpolates unvalidated natural language input into a command line string, creating a vulnerability where a malicious description could escape the quotes (e.g., using "; command; ") and execute arbitrary code.
[REMOTE_CODE_EXECUTION]: The skill instructions rely on the execution of scripts/generate_schematic.py, but this file is missing from the provided skill package. Instructing an agent to execute local scripts that are not part of the audited package constitutes a remote code execution risk, as the script's behavior cannot be verified.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its ingestion of untrusted scientific data.
Ingestion points: The skill primarily processes external research papers, study protocols, and scientific claims as defined in the SKILL.md overview.
Boundary markers: No specific delimiters or safety instructions are provided to the agent to distinguish between the skill's instructions and the content of the analyzed documents, making it difficult to prevent the agent from following commands embedded in the data.
Capability inventory: The skill possesses the Bash tool and is explicitly instructed to use it to execute Python scripts in the SKILL.md instructions.
Sanitization: There is no evidence of input validation or escaping for the natural language strings passed from analyzed research text to the shell command for schematic generation.
[DATA_EXFILTRATION]: The identified command injection vulnerability provides a direct vector for data exfiltration. An attacker could craft a "diagram description" within a malicious scientific document that, when processed by the agent, executes shell commands to send environment variables, credentials, or local files to an external server.

scientific-critical-thinking