The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it retrieves and executes diagnostic instructions from files within the project it processes. If these files are modified by an attacker (e.g., via a malicious pull request), they could influence the agent to perform unauthorized actions.\n
Ingestion points: Configuration is read from <project>/.claude/healthcheck.md and executable queries are read from <project>/.claude/protocols/db-healthcheck.md.\n
Boundary markers: The instructions lack delimiters or warnings to the agent to disregard instructions that might be embedded in the configuration data.\n
Capability inventory: The skill has access to Bash (used for curl, psql, kubectl), browser automation tools, and local file access.\n
Sanitization: Specifically for database checks, the skill is instructed to use queries from the config file "verbatim" without validation.\n- [DYNAMIC_EXECUTION]: The skill includes functionality to dynamically execute SQL and PHP code at runtime based on the contents of localized configuration files. This is most prominent in the database and usage report protocols.\n- [EXTERNAL_DOWNLOADS]: The skill metadata contains an update command (npx skills update) that triggers the download and execution of a package from the npm registry.\n- [COMMAND_EXECUTION]: The skill executes various system commands including curl, kubectl, psql, and pkill to perform its diagnostic tasks, using parameters supplied by the project configuration.

project-healthcheck