The Agent Skills Directory

[COMMAND_EXECUTION]: The skill recommends executing commands with elevated privileges using sudo to modify file ownership. This is a common but high-privilege troubleshooting step.
Evidence in SKILL.md: sudo chown -R $(whoami) ~/.npm.
Evidence in references/debugging-strategies.md: sudo chown -R $(whoami) /usr/local/lib/node_modules.
[REMOTE_CODE_EXECUTION]: Promotes the execution of installation scripts directly from the vendor's GitHub repository.
Evidence in references/installation-guide.md: curl -fsSL https://raw.githubusercontent.com/j0kz/mcp-agents/main/install-all.sh | bash.
Evidence in references/installation-guide.md: irm https://raw.githubusercontent.com/j0kz/mcp-agents/main/install-all.ps1 | iex.
Note: These scripts originate from the vendor's official repository and are treated as vendor resources.
[EXTERNAL_DOWNLOADS]: Fetches installation scripts and packages from external sources like GitHub and the npm registry.
Evidence: npx @j0kz/mcp-agents@latest and references to github.com/j0kz/mcp-agents/.
[DATA_EXFILTRATION]: Accesses local configuration files that could potentially contain sensitive data or environment variables used by the tools.
Evidence: cat ~/.config/claude/mcp-servers-config.json in SKILL.md.
[PROMPT_INJECTION]: Presents an attack surface for indirect prompt injection via the processing of untrusted log file data during debugging activities.
Ingestion points: references/debugging-strategies.md suggests reading and tailing log files such as mcp-*.log.
Boundary markers: Absent.
Capability inventory: Shell command execution via npx, curl, and sudo.
Sanitization: Absent.

mcp-troubleshooting