Menu
Reads the company brain before acting and writes the decision back afterward, keeping shared memory live. Use as the default loop for any agent doing real work in the codebase.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Answers a question from the company brain with cited sources and an honest list of what is not documented yet. Use when someone asks a factual question about the company, its systems, people, decisions or incidents.
Records a decision and its rationale back into the company brain after it is made, with sources, and detects supersede and contradiction effects. Use immediately after an architectural or process decision is reached.
Triages what needs human attention now: stale durable knowledge, unresolved incidents and aging gaps, ranked by severity and age. Use at the start of a work session or on-call shift.
Finds who owns or is responsible for a service, decision or area by traversing the typed knowledge graph and entity index. Use when someone needs to know who to talk to about X.
Helps a newcomer learn an unfamiliar area of the company or codebase by mapping its key entities, owners and decisions from the brain. Use in the first days on a new team, service or domain.
Reviews the company as a system and flags operational risks: unowned services, single-person bus-factor, bottlenecks, broken handoffs and redundant tools, each citing the real nodes behind it. Use for a health check or before reorganizing.
| name | read-before-act |
| description | Reads the company brain before acting and writes the decision back afterward, keeping shared memory live. Use as the default loop for any agent doing real work in the codebase. |
| version | 0.1.0 |
| triggers | ["before you start","default workflow","read before acting","agent loop"] |
| tools | ["Vitrus:think","Vitrus:gap_report","Vitrus:record_decision"] |
The brain only compounds if agents read it first and write to it after.
Vitrus:think the task to pull relevant prior decisions, owners and constraints.Vitrus:gap_report to see what is undocumented before you rely on it.Vitrus:record_decision with the decision, rationale and sources. It flags any contradiction or staleness your decision created.Reading without writing back wastes the loop; the next agent (or you tomorrow) starts blind. Close it every time.