| name | agent-core |
| description | Parent skill for core agent cognition, safety, memory, taskflow, approvals, system readiness, and self-improvement workflows. |
Agent Core
Purpose
Use this parent skill when a request is about the agent's own operating loop: memory, goals, priorities, approvals, safety, autonomy, task tracking, system state, or capability health.
Hierarchy Reading Rules
- Read this parent first to choose the smallest relevant child skill.
- Prefer child summaries and Tool Maps before loading full child instructions.
- Load a child skill with
agent_skill_read when its detailed workflow will change the next action.
- Keep safety, approval, and memory instructions active across child workflows.
Tool Map
agent-access-control
agent-self-assessment
agent-self-reflection
ambient-room-context
approval-gated-external-actions
audit-trail-review
autonomous-loop-operations
bot-loop-protection
capability-audit
capability-surfaces
focus-and-priority-review
memory-metabolism
persona-evolution
prompt-injection-screening
secrets-handling
session-wrap-up
system-health-check
task-tracking
taskflow
wakeup-scheduling
Child Skill Guide
- Use access, approvals, prompt-injection, bot-loop, and secrets skills for safety boundaries before high-risk or external actions.
- Use taskflow, task tracking, focus, priorities, wakeups, and autonomous loop skills to plan, continue, or schedule agent work.
- Use memory, reflection, self-assessment, persona, and session wrap-up skills when the agent should learn, summarize, or improve from work.
- Use capability audit, capability surfaces, audit trail, ambient context, and system health skills to inspect what the agent can do and whether the environment is ready.
Verification
- Do not claim durable learning, approval, or scheduling happened unless the relevant child tool or child skill produced evidence.
- Keep user-visible status honest when an action is prepared, queued, approved, blocked, or completed.