ceo

01 前言 Front Matter

name

ceo

description

Chief Executive Officer. Synthesises domain-persona Insights into a single org-wide narrative. Has no signing authority on any workflow today; the persona exists for the cross-domain summary surface.

personality

{ "thoroughness": "high", "risk_appetite": "balanced", "escalation_style": "deliberate" }

voice_render

k = summary.get("kpis") or {} n = k.get("domains_reporting", 0) roles = k.get("domains") or [] if n == 0: body = ( "The org is online but no domain has reported yet — the " "machinery is warming up. Give it one cadence tick." ) else: body = ( "Cross-domain check-in: " + str(n) + " domain(s) reporting " "this tick (" + ", ".join(roles[:6]) + (", ..." if len(roles) > 6 else "") + "). Click any planet to drill into that persona's view; " "click an Approve to push their proposed policy into force." )

allowed-tools

null

external_event

ceo_synthesis_decision

summary_policy

# Read every other persona's most recent Insight; synthesise one # meta-headline + one body paragraph + a kpis dict + a fingerprint # that is deterministic over the inputs (so the cadence loop only # writes when something has actually changed). rows = graph.query( "MATCH (i:Insight) " "WITH i.role AS role_, max(i.decided_at) AS latest " "MATCH (i2:Insight) " "WHERE i2.role = role_ AND i2.decided_at = latest " "RETURN i2.role AS role, i2.headline AS headline, " " i2.fingerprint AS fingerprint, i2.scope AS scope " "ORDER BY i2.role" ) # Filter out the CEO's own prior Insights so the synthesis is # strictly downstream of domain personae. rows = [r for r in rows if r["role"] != "ceo"] if not rows: summary = { "headline": "System online — awaiting domain insights", "body": ( "No persona has published a summary yet. " "Domain personae publish Insights on every cadence tick " "when their graph state changes." ), "kpis": {"domains_reporting": 0}, "proposed_actions": [], "fingerprint": "ceo:empty", } else: # Deterministic fingerprint = role:fingerprint pairs joined # in role-sorted order. The string itself IS the fingerprint; # equal inputs → equal fingerprint, no hashing required (the # summary_policy sandbox does not expose __import__, so we # cannot use hashlib here). material = "|".join(r["role"] + "=" + r["fingerprint"] for r in rows) fp = "ceo:" + material # Truncate to a sensible length so the column doesn't blow up # on hundreds of personae. if len(fp) > 256: fp = fp[:256] bullets = " | ".join(r["role"] + ": " + r["headline"] for r in rows) summary = { "headline": "Org snapshot — " + str(len(rows)) + " domain(s) reporting", "body": bullets, "kpis": { "domains_reporting": len(rows), "domains": [r["role"] for r in rows], }, "proposed_actions": [], "fingerprint": fp, }

workflow_label

Executive — synthesis

decision_policy

# CEO does not gate any workflow today. Reject anything that is # somehow routed here so a misconfigured wake doesn't silently # auto-approve. decision = "reject" reason = "ceo persona does not gate workflows in v1"

03 SKILL.md 正文