Read AGENTS.md first; it is the canonical repo context for coding
agents in this project.
Critical runtime invariant: Engineering Pulse is an agent- and skill-driven product.
Follow skills/engineering-pulse/SKILL.md and its references for the daily dashboard
workflow. Keep make run and scheduled runs executing the selected agent through
AGENT_CLI; do not replace that product path with direct Python orchestration.
Python files under scripts/ are tools called by the agent-guided workflow.
scripts/run_daily_dashboard.py is useful for helper/debug execution, not the
primary runtime.
For headless Claude Code runs, the generated runner must use explicit
non-interactive permission handling so Bash and file-read tools do not wait for an
approval UI that is unavailable during make run or LaunchAgent execution.