The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/report_received_today.py uses subprocess.run to execute the local dws binary. This is the intended behavior as defined in the skill's prerequisites and YAML frontmatter.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (report titles and content) received via the dws report inbox list and dws report entry get commands. While this data is rendered in Markdown tables and could theoretically contain injection attempts from other users, the risk is mitigated by standard LLM guardrails.
Ingestion points: scripts/report_received_today.py parses JSON output from tool executions.
Boundary markers: Not explicitly defined for the ingested report data.
Capability inventory: Execution of the dws binary via subprocess.run and writing temporary Markdown/JSON files in the current working directory.
Sanitization: The skill relies on structured JSON parsing; however, it does not explicitly sanitize the text content of reports before displaying them to the user.

dingtalk-report