The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute local Python scripts (knowledge-binding.py, get-market-config.py, run-metrics.py) and shell utilities (mkdir, sed, date) for environment setup, project directory management, and metric recording. All commands are scoped to the skill's root and project directories.
[EXTERNAL_DOWNLOADS]: The skill performs web searches to ground research topics in existing information. This is a primary function of the research pipeline and is used to avoid generating sub-questions for topics with no searchable data.
[PROMPT_INJECTION]: The skill ingests untrusted user input through the --topic parameter and processes external data from web searches. While this represents an indirect prompt injection surface, the skill implements proper slugification of user input for file paths and uses the data strictly for thematic decomposition.
Ingestion points: User-provided --topic parameter; search results from WebSearch snippets.
Boundary markers: None explicitly defined for search data interpolation.
Capability inventory: Write access to the local project metadata directory; Bash execution of internal scripts.
Sanitization: Implements strict alphanumeric slugification for topic-derived file paths to prevent command or path injection.

knowledge-plan