The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs several shell-based operations to manage files and execute external logic.\n
Evidence: Invokes a local Python script scripts/harvest.py from the associated mission-log skill to gather session data.\n
Evidence: Utilizes system commands like mv, wc, and grep to archive old logs and verify the integrity of new reports.\n
Evidence: Provides explicit instructions for users to set up automated execution through launchd or cron, creating persistent background tasks.\n- [COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary commands based on local configuration.\n
Evidence: It is designed to read ~/.claude/journal/config.env and execute any command defined in the JOURNAL_POST_CMD variable, passing the report path as an argument.\n- [DATA_EXFILTRATION]: The skill encourages the transfer of local activity logs to external platforms.\n
Evidence: Recommends synchronizing raw journal data across different machines using ssh or git repositories.\n
Evidence: Supports sending processed summaries to external notification services such as Notion, Telegram, or Slack.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its data ingestion patterns.\n
Ingestion points: Reads conversation data via harvest.py, as well as event data from calendar and task management CLI tools.\n
Boundary markers: Absent. The skill instructions do not specify any delimiters or warnings to prevent the agent from following instructions embedded in the gathered data.\n
Capability inventory: The skill has access to shell execution, file system modification, and arbitrary command execution via user-defined configuration.\n
Sanitization: Absent. There is no evidence of input validation or escaping for the external data included in the final Markdown reports.

daily-debrief