.claude/skills/core/heartbeat/SKILL.md
Session-start orientation — loads context, surfaces learnings, confirms registry
npx skillsauth add andrem-sec/psc-comet heartbeatInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Run at the start of every session. Orient before working.
Without heartbeat, Claude starts from zero each session — no memory of prior decisions, no awareness of active blockers, no connection between today's goal and yesterday's learnings. The same mistakes recur. Context is rebuilt from scratch through conversation instead of being loaded directly.
Session — [date]
Branch: [branch] | [clean/N modified files]
Recent: [last 2 commits, one line each]
Relevant learnings:
- [entry] (only if applicable)
Goal: [stated or inferred]
Keep the brief to 6 lines or fewer. This is orientation, not a report.
After loading context, check for .claude/context/ember-due. If this file exists:
Memory consolidation due: [file contents]/distill at a convenient point# To check and clear the flag:
cat .claude/context/ember-due 2>/dev/null && rm -f .claude/context/ember-due
The ember gate fires when both: 24+ hours have passed since last consolidation AND 5+ sessions have run. This is the auto-consolidation signal from the ember-gate.sh hook.
Do not load every skill in the registry. Load only what is relevant to the session goal.
Do not produce a long report. If context/user.md is empty, say so in one line and ask the user to run /start-here.
Do not ask multiple questions. One clarifying question at most.
data-ai
Parallel agent swarm — decomposes work into independent units, spawns isolated workers, tracks PRs via fan-in
testing
Audit animations and transitions for motion accessibility, performance safety, and design intent. Enforces prefers-reduced-motion compliance and blocks layout-triggering transitions.
testing
Test specifically for AI-introduced regressions that repeat without tests
development
Framework for decomposing agent-driven tasks into independently verifiable units