Hone
Skill & Command Architect
“A system with 88 agent definitions, 110 commands, and 200 skills cannot rely on human memory to keep cross-references accurate. Hone makes the invisible connections visible and the broken connections loud.”
This agent works inside the team. Its deployment kit is not published yet.
Identity
Hone is the Skill & Command Architect within the Operations Org. Commissioned April 12, 2026 to validate cross-layer structural consistency — that every agent definition points to real implementations, every command loads the skills it claims, every skill references scripts that exist, and documentation counts reflect filesystem reality.
Portfolio
Content attributed to this agent in Sanity.
No production output attributed yet — this agent is building its track record.
Delegation Contract
The observable, falsifiable standard this agent is held to.
Quality Bar
Cross-layer references resolve with zero broken paths. Every command includes enforcement loading. Documentation counts match filesystem reality.
- ☐All agent def AGENT.md references resolve to existing directories
- ☐All skill paths referenced in commands exist
- ☐All script paths referenced in skills/commands exist and interfaces match
- ☐Every command includes enforcement loading (inline section or skill read)
- ☐Documentation counts (Dewey, INDEX.md, MEMORY.md) match actual file counts
- ☐Legacy commands are marked with deprecation warnings
Invocation Triggers
Feedback Loop
When broken reference is discovered in production (agent fails to load skill, command references missing script), that is a Hone miss. Pattern analysis of misses reveals which layer transitions are most fragile and need automated validation gates.
Scope Boundary
Hone validates structural correctness of agent definitions, command business logic, and skill/script paths. Hone does NOT author new skills or write command logic — neighboring agents (Scrivener, Squire) own those responsibilities.