The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local shell commands for file management (mv), validation (wc, grep), and log harvesting (scripts/harvest.py). It also executes a user-defined command hook (JOURNAL_POST_CMD) retrieved from a local configuration file (~/.claude/journal/config.env). User-supplied date arguments are passed to shell commands, which may present a command injection surface if the agent does not properly sanitize the input.
[DATA_EXFILTRATION]: The skill processes session transcripts and report data, with instructions to optionally send this data to external notification services (e.g., Slack, Telegram, Notion) via a user-configured command hook. This functionality is part of the skill's primary purpose for automated reporting.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes session transcripts (jsonl files) which are untrusted data sources. Malicious instructions embedded in a user's previous chat sessions could influence the report generation or subsequent automated actions.
Ingestion points: Session logs in journal/raw/ and previously generated report files.
Boundary markers: The skill does not define the use of delimiters or 'ignore' instructions when interpolating transcript data into the report template.
Capability inventory: The skill can execute shell commands, write to the local file system, and access local configuration files containing potential environment variables or hooks.
Sanitization: No explicit sanitization or filtering of external transcript content is mentioned.

daily-debrief