The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a bundled shell script prepare-debrief.sh and the system open command to manage and view generated reports. The shell script performs directory creation using mkdir -p, but implements strict regex validation to ensure that the project slug only contains safe kebab-case characters, preventing command injection or path traversal.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted evidence from the current task (such as findings, metrics, decisions, and code diffs) to populate its generated HTML artifacts.\n
Ingestion points: Findings and metrics are pulled from the current task transcript and tool evidence as specified in the SKILL.md workflow.\n
Boundary markers: None; the skill does not define delimiters to separate untrusted task data from the HTML code structure.\n
Capability inventory: The skill can write files to the local project directory and launch them in the default browser using the open command.\n
Sanitization: The instructions lack guidance for sanitizing or escaping task evidence before it is embedded into the generated HTML/JavaScript environment.\n- [DYNAMIC_EXECUTION]: The skill generates self-contained HTML files containing inline JavaScript and CSS based on local templates and task data, then triggers their execution by opening them in a web browser.

html-playground