The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/briefing_ctl.py utility invokes the briefing_run.py script via subprocess.run. It uses a list of arguments for execution, which is a secure practice that mitigates shell command injection risks.\n- [COMMAND_EXECUTION]: In scripts/briefing_ctl.py, the cmd_history function builds file paths using user-provided domain names without explicit validation. This could allow for limited path traversal when attempting to read history files, though the impact is restricted by a fixed file suffix.\n- [PROMPT_INJECTION]: The skill architecture is susceptible to indirect prompt injection because it ingests data from external RSS feeds and APIs.\n
Ingestion points: Untrusted data from external sources is collected and stored in domains/<name>/data/today_context.json.\n
Boundary markers: The provided prompt templates (generate_base.md, review_base.md) do not include explicit delimiters or instructions to ignore embedded commands.\n
Capability inventory: The pipeline can execute shell commands via subprocesses and send emails to configured recipients.\n
Sanitization: There is no evidence of sanitization or validation of the external content before it is interpolated into LLM prompts.

briefing