The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's instructions in SKILL.md use bash shell commands (test -f) to detect the existence of local preference files (EXTEND.md) to customize its behavior.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface typical of data-processing agents.
Ingestion points: Untrusted data enters the agent context through search results retrieved via kimi_search_v2 as instructed in SKILL.md.
Boundary markers: The skill does not employ specific delimiters or instruction-ignore warnings for the searched content.
Capability inventory: The skill has the capability to write files to the local file system using scripts/generate_dashboard.py within the PythonRun environment.
Sanitization: The Python script performs direct interpolation of search data into HTML templates using f-strings; while it uses json.dumps for structured data, many fields are inserted without HTML entity encoding, representing a potential Cross-Site Scripting (XSS) surface in the generated local dashboards.

canghe-tianyancha