The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute several system commands including uname, python3 --version, Rscript --version, node --version, git --version, claude --version, and claude mcp list. These executions are limited to version discovery and environment verification, consistent with the skill's stated purpose as a diagnostic tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from local command outputs (such as tool version strings and the list of connected MCP servers) to generate a diagnostic table, creating a vulnerability surface where maliciously crafted tool output could influence the agent.
Ingestion points: Command outputs from environment discovery tools and claude mcp list (found in SKILL.md and references/setup-checklist.md).
Boundary markers: The instructions provide specific version comparison and parsing rules in references/setup-checklist.md to structure how the agent interprets the ingested data.
Capability inventory: The skill utilizes Bash for read-only system queries and writes a local diagnostic log to ~/.medsci-skills/ (SKILL.md).
Sanitization: The instructions include strict anti-hallucination rules, requiring the agent to report exact stdout and failure messages without fabrication (SKILL.md).

setup-medsci