The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's integration wizard workflow involves the agent reading and analyzing project source files and dependencies to plan and implement tracking. This ingestion of untrusted code creates a potential surface for indirect prompt injection.\n
Ingestion points: Project source files, node_modules directory, and the generated .posthog-events.json file.\n
Capability inventory: Source code analysis, file editing (to insert tracking calls), and interaction with the PostHog platform via MCP tools.\n
Boundary markers: The instructions do not specify the use of delimiters to isolate untrusted project content from the agent's internal reasoning or operating instructions.\n
Sanitization: No explicit filtering or sanitization of project content is mentioned prior to analysis.\n- [DATA_EXFILTRATION]: The provided Ruby example project includes functionality to capture application exceptions and transmit them to PostHog's servers. This involves sending data such as exception types, messages, and stack traces (containing file paths and line numbers).\n
Evidence: Manual exception capture in todo.rb sends $exception events with type, message, and backtrace details to the PostHog API.\n- [COMMAND_EXECUTION]: The skill documentation and example project README instruct the user to perform standard Ruby package management tasks to set up the environment.\n
Evidence: Instructions for gem install bundler and bundle install in references/EXAMPLE.md.

integration-ruby