Financial Close

Agents

Seven agents watch this module. They brief, advise, order and propose cover — the entity rollup, the dependency graph, the flux thresholds, the skill graph and the lock conditions are all computed before any of them is called. No agent here can change a task status, reassign work, or lock a period.

Replayed proposals

Agents in this module

2 model tier(s)

Calls this run

Taken from the audit trail

Mode

Scripted

Fixture-backed and deterministic

Figures an agent originated

Enforced by the numeric guardrail

The roster and its boundaries

What each agent does, what it cannot do, and the code that makes that true

Agent	Does	Never does	Bounded by	Calls
C1Close Risk Agent claude-opus-5 (replayed)	Briefs the controller on where the close stands, what is holding it up and who must act.	Changes a task status, a critical path or a lock condition.	`close/tasks.ts and close/lock.ts compute the state it describes`	1 0 ms
C2Flux Commentary Agent claude-sonnet-5 (replayed)	Writes year-on-year flux commentary and states an evidence gap where the data does not establish a driver.	Decides which movements need explaining — thresholds do that before it is called.	`close/flux.ts applies the policy thresholds`	1 2 ms
C3Critical Path Agent claude-opus-5 (replayed)	Recommends the resequencing or reassignment that would pull the close date in, and names what each move depends on.	Reorders the graph. The dependency structure and the longest remaining chain are computed, not advised.	`close/tasks.ts — orchestrate() derives the critical path`	1 0 ms
C4Capacity & SLA Agent claude-sonnet-5 (replayed)	Reads owner load against working capacity and says where the close is capacity-constrained rather than process-constrained.	Reassigns a task or changes a due day.	`close/skills.ts computes open load, capacity and utilisation per person`	1 0 ms
C6Task Reassignment Agent claude-opus-5 (replayed)	Proposes who should take over a task that is holding others up, citing the work that person has actually completed.	Names anyone the evidence does not qualify, or performs the reassignment. Eligibility is decided before it is called and the handoff is a human action.	`close/assignment.ts filters on entity authority, role and completed-task evidence; close/skills.ts builds that evidence from history`	1 0 ms
C7Entity Readiness Agent claude-sonnet-5 (replayed)	Reads the group rollup and says which of the twelve entities will not close, and what is stopping each one.	Changes an entity status. Status is derived from that entity’s own tasks and cannot be reported green over a late one.	`close/entities.ts — rollUpClose() derives every status from the task set`	1 0 ms
C5Lock Readiness Agent claude-sonnet-5 (replayed)	Puts the unmet lock conditions in the order they should be cleared and names the owner of each.	Locks the period or marks a condition met. The lock refuses in code, not in prose.	`close/lock.ts — assessLock() decides; actions.ts refuses`	1 0 ms

Deep reasoning

Judgement-heavy work where the cost of a wrong call is high.

claude-opus-5

C1C3C6

Workhorse

The default. Structured proposals with a rationale.

claude-sonnet-5

C2C4C7C5

What each agent did this run

Newest last, straight from the audit trail

#8
C2 Flux Commentary Agent
9:14:26 am
Drafted flux commentary on 26 movements
claude-sonnet-5 (replayed)2 ms
#9
C1 Close Risk Agent
9:14:26 am
Drafted the close risk briefing
claude-opus-5 (replayed)0 ms
#10
C3 Critical Path Agent
9:14:26 am
Advised on the critical path
claude-opus-5 (replayed)0 ms
#11
C4 Capacity & SLA Agent
9:14:26 am
Assessed close team capacity against proven skills
claude-sonnet-5 (replayed)0 ms
#12
C7 Entity Readiness Agent
9:14:26 am
Assessed readiness across 12 entities
claude-sonnet-5 (replayed)0 ms
#13
C6 Task Reassignment Agent
9:14:26 am
Proposed cover for 2 stranded task(s)
claude-opus-5 (replayed)0 ms
#14
C5 Lock Readiness Agent
9:14:26 am
Ordered the outstanding path to a period lock
claude-sonnet-5 (replayed)0 ms