synthesis-daily-rituals

Warn

Audited by Gen Agent Trust Hub on Sep 24, 2026

Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [PERSISTENCE]: The scripts/install_day_end.py script installs a macOS LaunchAgent (com.synthesis.day-end-nudge.plist) into the ~/Library/LaunchAgents/ directory to provide periodic reminders, establishing persistence on the host system.
  • [COMMAND_EXECUTION]: The skill checklists command the agent to run various shell utilities and Python scripts for project auditing and environment verification. These include git, gh (GitHub CLI), launchctl, and system commands like date and osascript for notifications.
  • [INDIRECT_PROMPT_INJECTION]: The ritual workflow ingests untrusted content from Slack, Google Chat, and email to ground its message drafting and context updates. This constitutes a surface for indirect prompt injection as specified in category 8.
  • Ingestion points: Slack transcripts (transcripts/slack/), Google Chat transcripts (transcripts/gchat/), and email logs in the workspace repository.
  • Boundary markers: The skill specifies a 'Pre-Send Review Gate' and 'Draft Grounding' rules to warn the user, though delimiters for ingested data are not explicitly defined in the provided scripts.
  • Capability inventory: The skill possesses extensive capabilities including file writing (atomic_write), shell command execution (subprocess.run), and network operations (via gh and potential MCP tools).
  • Sanitization: No explicit sanitization or escaping of the ingested chat/email data is performed before it is presented to the agent for draft generation.
  • [DYNAMIC_EXECUTION]: The installer (install_day_end.py) dynamically writes executable Python and Bash scripts to ~/.synthesis/day-end/bin/ and creates a symlink in the user's local binary path. Additionally, scripts/day-end-nudge.sh executes dynamically generated Python code via stdin.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 24, 2026, 12:51 AM