@integrity-labs/agt-cli 0.28.472 → 0.28.474
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/bin/agt.js +3 -3
- package/dist/{chunk-KCU4ZTE4.js → chunk-C47UJ5QI.js} +2 -2
- package/dist/lib/manager-worker.js +94 -9
- package/dist/lib/manager-worker.js.map +1 -1
- package/dist/mcp/augmented-admin.js +1 -1
- package/package.json +1 -1
- /package/dist/{chunk-KCU4ZTE4.js.map → chunk-C47UJ5QI.js.map} +0 -0
|
@@ -21781,7 +21781,7 @@ server.tool(
|
|
|
21781
21781
|
);
|
|
21782
21782
|
server.tool(
|
|
21783
21783
|
"debug_query_audit_log",
|
|
21784
|
-
'Read recent audit events to troubleshoot why/when agents bounced \u2014 primarily restart events. Defaults to the "agent.restart" action: each event returns { created_at, source, reason, actor_kind, agent_id, code_name }, where `source` is the restart trigger (agent-self-request, channel-command, maintenance-window, stale-mcp, hot-reload-mcp, mcp-presence-reaper, day-rollover, auth-tuple-change, \u2026) and `reason` is the free-text rationale when one was given (e.g. the agent\'s own words for a request_restart). Two modes: (1) pass { agent_id } to scope to ONE agent (fails closed if you\'re not authorized for its org) \u2014 use when a specific agent went quiet. (2) OMIT agent_id for a CROSS-AGENT sweep across every org you\'re authorized for (optionally narrow with { org_id }) \u2014 use to spot fleet-wide restart storms or compare agents. Filter the action with `action` (pass "all" to drop the action filter and read every event type), and bound with { since_hours?, limit? }. Cross-agent results are org-walled in SQL; unauthorized orgs are invisible.',
|
|
21784
|
+
'Read recent audit events to troubleshoot why/when agents bounced \u2014 primarily restart events. Defaults to the "agent.restart" action: each event returns { created_at, source, reason, detail, actor_kind, agent_id, code_name }, where `source` is the restart trigger (agent-self-request, channel-command, maintenance-window, stale-mcp, hot-reload-mcp, mcp-presence-reaper, day-rollover, auth-tuple-change, \u2026) and `reason` is the free-text rationale when one was given (e.g. the agent\'s own words for a request_restart). `detail` (ENG-8293) is the structured "what changed?" payload, null for sources that don\'t emit one. For `auth-tuple-change` it is { auth_tuple: { changed[], unchanged[], components[] } } naming WHICH credential/identity component moved and its before/after fingerprint \u2014 the auth tuple is deliberately NOT keyed on the rotating access token, so an auth-tuple-change is always a genuine credential or identity change, and `detail.auth_tuple.changed` is how you tell which one. Comparing `components[].to` across agents answers whether a restart WAVE was one control-plane push (same new value everywhere) or coincidental per-agent rotations. Two modes: (1) pass { agent_id } to scope to ONE agent (fails closed if you\'re not authorized for its org) \u2014 use when a specific agent went quiet. (2) OMIT agent_id for a CROSS-AGENT sweep across every org you\'re authorized for (optionally narrow with { org_id }) \u2014 use to spot fleet-wide restart storms or compare agents. Filter the action with `action` (pass "all" to drop the action filter and read every event type), and bound with { since_hours?, limit? }. Cross-agent results are org-walled in SQL; unauthorized orgs are invisible.',
|
|
21785
21785
|
queryAuditLogSchema.shape,
|
|
21786
21786
|
async (args) => {
|
|
21787
21787
|
try {
|
package/package.json
CHANGED
|
File without changes
|