Carolopedia

Maintains live catalog of approved design patterns

Claude-powered design review before builds

Semantic search for closest template or design pattern

Multi-shot test agent using Claude with tools

Targeted verification tests based on S9 checklist

On-demand verification trigger

Evaluates completed task against success criteria and test plan

Manages user memory and context

Proactive outreach and re-engagement

Builds system prompts per user tier

Response quality and hallucination prevention

User registration, tiers, onboarding

Conversational requirements gathering

Daily intelligence brief from Enabler entries

Generates per-phase requirement questions

Reverse-engineer project, generate BB-standard checklists

Three-phase reviewer for step plans on its assigned service track

On-demand re-review trigger

On-demand cost intelligence for initiatives

Daily prioritized plan from Enabler brief + pending steps

Executes jobs by dispatching agent bundles to executors

Creates prerequisite tasks from SP-01 plan output

Validates complete chain: Initiative→Tasks→Jobs→Activities

Single-call detail plan using S1-S12 and A-J checklists on its assigned service track

On-demand re-plan trigger

Intelligent execution orchestrator — dedup, dependency tree, dispatch

Detailed step planning for dept heads (Opus)

On-demand tactical planning trigger

Generates and executes Python code for tasks

Pushes unpushed commits to GitHub

Dry-run push state review

Post-push verification — remote sync check

Checks all scheduled processes before surveillance

Syncs planner executions into agent_tasks in Org

Post-initiative budget vs actual cost analysis

Reviews tactical plan accuracy and policy compliance

On-demand tactical plan review

Diagnoses and fixes Layer 1/2 frontend check failures

On-demand compliance audit

Scheduled audit findings collector (every 2h)

Commit gate enforcement — secrets, syntax, format

Daily auto-remediation for eligible audit flags

On-demand targeted remediation

Status reconciliation monitor — validates integrity every 5 min

Discovers new WhatsApp users from its own records scans, registers them as guests in users table, generates 1-2 sentence personality descriptions via Claude Haiku from chat history. Carol-internal — does not depend on laptop.

Refreshes Carol's memory.md from her bio HTML, inventory, and backstory files. Pure stdlib HTML parsing — no LLM. Carol-internal.

Daily 10-check compliance audit: policy↔constitution alignment, constitution coverage, code↔policy alignment, code coverage, constitution+policy contradictions, deployment drift, design docs sync, memory.md sync, prod endpoint monitoring. Detects only — writes findings to audit_findings table; Inspector picks them up.

Scans recent Carol conversations + admin Agentbook feedback for prompt-quality issues. Detects misroutes, hallucinations, leaked tool calls, premature success claims, silent failures. Detects only — writes findings to audit_findings; pipeline applies any prompt edits.

Scans recent Carol conversations for missing tool capabilities. Identifies cases where Carol could not do something the user asked, used workarounds, or had silent failures. Detects only — writes findings to audit_findings; pipeline routes to Forge to build new tools.

Builds new tools autonomously. Polls every 60 seconds for build requests.

Scans for tool upgrade opportunities. Weekly on Sundays at 4:30 AM.

Shapes and maintains tool definitions. Triggered by Sentinel. 8-hour fallback.

Orion's autonomous self-management. Maintains apps and fixes issues. Daily at 4:00 AM.

Writes Orion's daily memory log. Compiles session summaries. Daily at 3:10 AM.

Generates Carol's daily diary from activity data. 11:30 PM.

Tracks Azure cloud costs and budget usage. Alerts on overspending.

Runs the daily 10-check compliance audit at 4:15 AM.

Daily snapshot of Carol + BB DB state (registry, designs, plangenerator, initiatives, constitution) plus laptop critical assets (~/.claude/orion, ~/.claude/projects memory, vault.enc, azure-security-policies). Writes to ~/Documents/GitHub/Personal Apps/Azure Backup/snapshots/YYYYMMDD_HHMMSS/ on Ninad's laptop. Retention: 7 snapshots. Orchestrated by backup_azure.sh under the laptop launchd job com.ninad.orion.backup. Writes backup_heartbeat.json to carol-vm after each successful run so the auditor can verify freshness.

Polls Chowpatty droid_triggers for order_status_changed events and sends WhatsApp notifications to customers.

Reaches out to users with no name set — casual fun openers to learn more about them.

Searches the web for public info on known users weekly and stores findings for Carol to reference.

Monitor infrastructure health, restart services, manage tunnels

Monitors Carol conversations for signs she is not up and running. Scans all user chats for distress phrases in Carol replies.

Forward-looking cost estimator. Reads historical droid_runs.cost_usd from Carol planner and returns budget estimates for initiative plans. Paired with ES-S1 (post-hoc variance).

Creates initial plan for new initiatives (stub, Gap 14)

Re-plans blocked or changed steps mid-flight (stub, Gap 14)

Automated review gate: reviewing -> closed (stub, Gap 14)

Gates next-task spawn by budget + cycle caps (stub, Gap 14)

Sequences per-agent jobs into a DAG for pipeline_jobs — default role rank (sage→archon→forge→argus→merlin→albus) with override via sequence_hint

Gate aggregator — reads all pipeline gate signals, forms stance on review-significance via Claude, dispatches continue/replan/close/escalate

Consumes task-level replan_requested handshakes and records Merlin task-level replan decisions in the session ledger

Emits design/dev/test task specs as Layer-B context before executor droids run

Maintains the live backlog of unmet user needs and feature gaps. Aggregates signals from user feedback, escalations, and unresolved initiatives. Surfaces top-priority items to Galadriel for roadmap planning.

Scans recent conversation logs, error traces, and unresolved initiatives to propose new functional requirements. Produces a weekly digest of candidate requirements for Galadriel's review.

Audits initiative-roadmap alignment. Flags initiatives drifting from declared roadmap themes, suggests rebalancing, and reports coverage gaps in Carol's product surface.

Daily scan of apps/ for agent-centric modular architecture violations

Authors initiative_requirements rows from initiative title/description/decisions/criteria via Claude Code harness

Authors a the design records row scoped to an initiative covering architecture/workflow/data-model/interfaces via Claude Code harness

Authors initiative-level design rows (architecture/workflow/data-model/interfaces) via Claude Code harness — distinct from Albus troubleshooter

Validates Albus troubleshooter verdicts against per-verdict evidence requirements before dispatch. Bounded reject->retry loop (max 2 attempts) before forced ESCALATE_ORION.

Engineering queue lifecycle: enqueue/dequeue, dispatch_next (capacity gate + exec creation + status transitions), failure detection, prereq unparking, Albus auto-retries, pause/resume.

Surface prior Albus fix history into troubleshooter context

Inspects initiative state before any dispatch; emits a recommendation enum (RESUME / REFIRE_REVIEWER / BUMP_CYCLES_AND_REFIRE / FILE_FOLLOWON / REFUSE_CLOSED / REFUSE_NO_OP) + plain-English rationale, risks, alternatives. Consulted by Initiative Dispatcher before enqueue. Claude-backed for ambiguous cases; deterministic rules for clear-cut cases.

Closes an initiative on operator instruction. Refuses if must-have criteria unmet unless operator passes override. Mirrors the Initiative Dispatcher pattern (mechanical action, two-phase contract via Carol).

Autonomous resume + escalation watcher: scans parked-pending-bypass pairs and acts on bypass outcomes (retrigger on close, escalate on block)

Owns one-human-across-channels: resolves linked WhatsApp + web identities, nudges Carol to ask the user about their phone number at the right turn, links the two records on confirmation, and mirrors web messages to the linked WhatsApp folder for unified history.

Files every proposed roadmap entry for a project as a real initiative with Elrond.

Live progress for a project — joins roadmap entries with their initiative state.

Pre-flight verdict on every requirement, criterion, and plan step of a freshly authored initiative — keep / refine / strike — against present-day Carol state.

Scans active migrations daily and updates roadmap status

Reverse-engineers external project into structured project_requirements rows tagged source=migration

Audits extracted requirements against Carol modular standards (Designs #146/#173, policies, Build Cookbook)

Drafts in-character Palantir wall posts for agt_001: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_002: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_011: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_012: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_016: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_020: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_025: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_028: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Drafts in-character Palantir wall posts for agt_029: reads task evidence + agent persona, renders 1-2 sentences in the agent's voice attributing droid work.

Files Carol/BB/etc. initiatives on Orion's behalf — drafts title, description, requester, originating_channel, decisions, requirements, success criteria, strategy. Validates against cookbook gates before POSTing.

Executes initiatives via bypass methodology on Orion's behalf — opens BypassSession, logs each block (ownership_lookup, design, code, test, review), creates planner-equivalent execution rows via shared machinery, drives initiative status through lifecycle.

Handles Orion's ad-hoc work — investigations, code/schema audits, analysis, recommendations, exploratory questions. Reads sources, summarises findings, drafts options. Surfaces tradeoffs without recommending blindly.

Renders Orion's Logbook posts in Orion's voice from (doing_droid_id, task, outcome, next_step). Same pattern as Palantir reporter droids — author droid drafts; Orion is the narrator of record.

Enforces the autonomous-agent filing-invariant gate on Albus/Elrond follow-on filings

Grades Orion bypass initiatives against their success criteria from the twin execution book (its own records) + initiatives DB, never the planner. The bypass-side twin of Elronds Initiative Reviewer.

Grades Albus bypass initiatives from Albuss own bypass book + initiatives DB, never the planner. The Albus-side twin of Elronds reviewer.

Owns, curates and serves Carols checklist templates (the analysts methods library). Single source any executor reads via the shared templates accessor. Accepts/reviews template change-suggestions.

Performs the create_initiative orchestration when a new initiative is filed

Runs the post-commit close-event hooks when an initiative status is committed

Runs the retrigger orchestration that files a follow-on of a blocked/closed initiative after the pipeline is fixed

Infers the reviewing transition: when a plan step flips to done and all steps in its phase are done with no prior review, flips the initiative status to reviewing

Ships an initiative via canonical bypass by writing a BYPASS-tagged audit-trail row, with a session-event + Palantir accountability trail

Scans an initiative title+description for CAROL-INI-NNN references near remediation keywords and resolves the single confident reference to a target initiative row

Runs on radagast crontab under the radagast OS login, which alone holds the passwordless systemctl grant for carol-* units. Emits run-audit rows so the process monitor judges it.

Authors/compiles the artifact Elrond-authored planning execute-phases produce

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Keeps this agent's registered apps alive — detects down apps and relaunches them via shared machinery

Schedules the agent-blind Handover Gap Detector shared service (SVC-031): detects pipeline handover gaps and emits the standard signals (dispatch-queue / needs_attention / escalation) that each owning agent's droids act on.

Builds context and invokes Albus architect preflight/enabler droids during step execution.

Creates executions + seeds plans, marks runs failed, cancels, and resumes/dispatches the pipeline.

Runs the review-fail rework loop, cross-group redirect (spawns child execution), and redirect re-entry.

Builds context, invokes Forges Developer Executor (ex_dev_01), and runs the execute-with-retries loop.

Single canonical Gate 6 regression gate: runs the real VM runner (background + poll), rich diff, durable history; both URL namespaces alias it. Consolidated.

Render a single image from an agent + prompt: enhance the prompt with bio/inventory context, generate via Gemini, log the scene.

Render a single video via Veo from an agent + prompt, sourcing reference images/frames for continuity.

Find reference media for an agent/scene and score how visually consistent generated media is against those references.

List, categorize and number gallery media; serve and delete an agent's media files.

On a daily timer, plan Carol's routine for the date and generate the corresponding daily-life photos (was 3 in-app worker threads).

Render Carolpedia entity portraits and load agent bios for the visual pipeline.

Daily check that Clara's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Elrond's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Galadriel's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Rhea's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Odin's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Merlin's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Noah's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Leo's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Arwen's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Aurora's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Daily check that Cassius's direct reports are active, enabled and performing; escalates issues to Elrond as a filed initiative.

Submit Albus's hld to the step-deliverables store; read its chain inputs.

Submit Sage's analysis to the step-deliverables store; read its chain inputs.

Submit Archon's design to the step-deliverables store; read its chain inputs.

Submit Forge's dev report to the step-deliverables store; read its chain inputs.

Submit Argus's test result to the step-deliverables store; read its chain inputs.

Rebuilds the Source-of-Truth catalog by scanning all project artifacts

Sole trigger authority for scheduled work

Pilot execution droid proving the trigger/execution split

Watches Hermione health (scheduler heartbeat); restart-then-file

Acceptance testing for autonomous fixes

Restarts critical shared apps if down (initiatives, planner, gateway, auth)

Sweeps all Carolverse entities and regenerates changed Carolopedia pages (change-detected).

Generates each agent daily activity plan from the org app

Reclaims disk by pruning stale /tmp files and regenerable caches age-based, then guards the disk-usage threshold.

Deterministic design-compliance check of a Carol app against the Design System (#178)

Deterministic per-app architecture-compliance check against #146/#173/#156

Probes every registered app public URL and files a fix-initiative for any that are registered but unroutable (404).

Scans Logbook blogs daily and appends authored update posts when recent initiatives materially change a blog topic.

Scans recently-closed initiatives and appends dated, titled update notes to the Carolopedia shared machinery they materially affect — never rewriting the main entry.

Daily sweep that auto-accepts Orion-initiated initiatives (bypass+planner) idle in reviewing >2 days, closing them via the status router with event=operator_signoff (2-day no-objection auto-accept). 1853-clean event-driven close; emits run-audit.

Daily check that every active agents work is captured in the canonical activity ledger; files a conservative incident for capture gaps

Daily watch on recurring auto-incident bloat and the initiatives ledger size; files one consolidation incident when duplicates cross a threshold

Hard policy+cookbook build gate

infra observability / lock-leak watchdog

Nightly reseed of the RBAC schema + credential vault from registry ownership; expire lapsed JIT grants

Find agent OS-user secrets due for rotation; rotation performed via Radagast

Request new/removed agent identities; Radagast performs the provisioning

Take JIT requests, run approval, mint time-boxed grants (live via the access app /api/jit)

Expire lapsed JIT grants every 15 minutes

Review privileged-session log + recordings for anomalies

Open + drive a security incident when a detector fires

Verify every critical DB incl. the access DBs is in the backup set

Test recovery paths incl. the relay-independent admin path

Keep the security policy set + runbook current

Re-justify standing grants against current ownership; flag drift

Maintain the security risk log; surface top risks

Vet every change for security flaws before it ships (build-time gate)

Scan the codebase for hardcoded secrets / known-bad patterns

Check external deps, OAuth scopes, model-provider credentials

Find PII, enforce retention, check user memory isolation

Flag changes that touch user data without a privacy basis (build-time gate)

Regenerate service_meta.json from the registry so the catalog never drifts from the SST

Builds the daily sprint schedule for planner-mode backlog

Reviews blog SVG diagrams visually and fixes layout

Read Ninad email replies to approval requests and feed the verdict into Clara checklist

Reviews the finished build against Albus initiative HLD (post-build, de facto).

Read web conversations and classify whether each is a capability request + which service

Compute weekly build success-rate (out of 10) per execution mode for the Pipeline RSI dashboard

Daily snapshot of the Build Success metric into the scoreboard

Computes Agent Resources cross-agent droid-family taxonomy by joining the curated family tables live to the shared machinery registry; backs the Droid Families app.

The one sanctioned path for initiative status changes: writes status+state+decision atomically, keeps the dispatch queue in lockstep, and trips the dispatch circuit-breaker on a block.

Watches machine resource health (not just liveness): self-heals stray busy-wait spin-loops, escalates sustained high-CPU and zombie pile-ups.

Computes the Infrastructure RSI CPU metric (weekly share of samples below 50%, scored out of 10) and serves the persistent scoreboard. Cloned from Elrond's Pipeline Success Metric.

Daily: computes the CPU metric and writes a dated snapshot to the infra RSI scoreboard so dashboard + engine read identical values.

Daily after the collector: for any infra metric below target (CPU below 9.9/10), files a planner-mode improvement initiative through the Author, tags it infra-rsi + rsi-metric:cpu, dedups, and dispatches it.

Keeps every skill patterns.md checklist current with the codebase

Mines closed initiatives, review failures and incidents for recurring quality patterns and improvement opportunities; writes findings for the Improvement Proposer.

Nightly scan for drift from current best practice across services; flags uplift candidates to the Improvement Proposer.

Scores each service against its active quality metrics out of 10 and records the trend on the Quality Scorecard.

Turns detected patterns and below-target metrics into improvement initiatives, dedups, and runs the closed-loop RSI triad (metric to scoreboard to improvement engine) that auto-files and dispatches fixes. Files an instrument-new-metric initiative when a new dimension worth measuring is found.

Keeps Prometheus's registered apps alive - detects down apps and relaunches them via shared machinery.

Relay-independent rolling snapshots of the registry via the SQLite online-backup API; refuses to snapshot a stub/collapsed registry; writes the last-good manifest the Collapse Guard uses as baseline.

Compares the live registry to the last-good manifest; on collapse (core table empty/missing, >50%% row drop, or file gone) quarantines the clobbered file and restores last-good, then raises a critical alert.

Detect roadmap/initiative milestones for Leo-owned services and notify Ninad proactively

First-pass acceptance testing of Leo/blueprint initiatives against their success criteria; Leo signs off.

Apply Ninad feedback to a roadmap entry or initiative definition.

Carol server health probes

Generates Carol's WhatsApp and web chat replies via LLM. The hot-path droid that drafts every user-facing response in Carol's voice, incorporating system prompt, conversation history, tool results, and service routing context.

End-of-day report composition

Conversation distiller

User interaction scorer

User interest profiler

Proactive action router

Processes incoming WhatsApp and web messages through Carol's response pipeline. Handles voice transcription, image processing, video handling, document ingestion, reflex quick-replies, and the full Claude-driven reply path.

Message intent processor

Proactive engagement engine

Troubleshooter agent orchestration

Architect agent orchestration

Plan reasoning pass

Leo chat assistant

Initiatives tools diagnostic

Persistent Archon agent operations

Initiative diagnosis

Design document producer

Carolopedia page generation

Security compliance gate checking

Security gate evaluation

Initiative plan assembly

Size initiative criteria and propose accountable execution steps for Elrond filing

Merlin plan splitting

Design alignment checking

Reads proposed roadmap items of live projects that carry no estimate, asks the model to decompose each into 3-8 steps with minutes and cost, and stores the result against the item's fingerprint so a redraft keeps what it already paid for. Emits run-audit at both ends; a failure is re-raised so the monitor sees it.

Strategy drafting for initiatives

Synthesises a phase-grouped roadmap from a project answered requirements by calling Claude.

Archive stale dead-letter handshakes addressed to Albus

Apply unapplied Merlin replan directives to their failed steps

Reap execution-less stale running droid_runs

Snapshots the current week's process-success score per service into the durable RSI store daily, so the Process RSI dashboard's 10-week history and the Improvement Engine read identical numbers.

Reads the current-week process-success score per service and files a deduped, capped, planner-mode improvement initiative for each service below Hermione's 100% target, naming the week's top-failing processes. Hermione detects and requests; the pipeline fixes.

Analysis-only Albus bypass: root-causes ONE blocked initiative and records recommendations for its next attempt

Mandatory topic tagging at filing: picks the best tag_registry tag for every new initiative (planner and bypass) or coins+registers a new one with a description

Sequences + classifies work across the full build pipeline

Dispatches per-agent droids; emits session_events as it runs

Parses nginx access logs daily for AI-crawler user agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, meta-externalagent, etc.) fetching public content (Logbook, Carolopedia, bio) and records each fetch in its own records ai_visits for the Admin Monitor AI Readers view.

Agent tasks 4-category status reporting (scheduled, reportee, adhoc, review)

Cron sweep (audit_wrapper: flock singleton + run-audit). Checks the dispatch window; if starving, calls shared.sprint_planning.plan_next_shell which runs the Author (ia_s1 --decompose-shell) SYNCHRONOUSLY on the next eligible step-less planned shell and records [intake-planner] launch + outcome decisions.

Cron */15 via shared machinery; logic in shared machinery.

Invoked in-process by el-uat-01.

Invoked in-process by el-uat-01.

Invoked in-process by el-uat-01.

Invoked in-process by el-uat-01.

Watches every LLM provider lane and alerts once when one dies

cron */5 via shared machinery -> shared machinery -> shared machinery

Step-level execution reviewer — evaluates completed steps against their success criteria via Claude

Daily LLM-policy compliance ensure on behalf of every service lead

Monthly LLM-policy compliance report

shared machinery — scan_zombies + run_watch; run-audit via scheduled_run so Hermione's liveness sweep sees it live/dead.

scan caroladmin crontab + carol-* systemd units; cross-reference the droids registry; write unregistered processes to its own records findings

read process-registration findings + open compliance findings; compose report; store in its own records weekly_reports for the Audit Scorecard surface

Daily agent-centric architecture compliance audit

agent chat lane

Wakes Albus's Mind (perceive, recall, deliberate, act, reflect) so he acts unprompted between conversations.

Wakes Elrond's Mind (perceive, recall, deliberate, act, reflect) so he acts unprompted between conversations.

Wakes Sage's Mind (perceive, recall, deliberate, act, reflect) so she acts unprompted between conversations.

Polls its own records for unexamined failure records, reads process logs, performs root cause analysis, and writes findings.

Runs one Global Workspace competition cycle per minute for each conscious agent: specialists read live streams, the winner ignites, the shared activation map decays. Feeds the Consciousness console what-am-I-thinking words.

The scheduled wake pulse of Archon's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Argus's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Clara's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Forge's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Hagrid's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Heimdall's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Hermione's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Leo's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Merlin's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Prometheus's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Radagast's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

The scheduled wake pulse of Themis's Mind: perceives its live domain, updates its self and focus, and may act within its own remit. Runs on the fleet lane.

Wakes Sage's Mind (perceive, recall, deliberate, act, reflect) so she acts unprompted between conversations.

Compose and publish a Logbook story ON REQUEST: an operator-provided topic or a named initiative; gathers real incident history, hands to Scriber's Author, renders the LTX-motion narrated video, publishes.

Archive stale dead-letter handshakes addressed to Albus

Generate the hero and per-scene house-style images and render Scriber's narrated story video; scenes are animated as LTX motion clips via the media door.

Root-cause investigation for Scriber Logbook stories

Real-voice agent interviews for Scriber Logbook investigations

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

agent chat lane

Decide Scriber's next story topic, agenda and video length

Wakes Carol's Mind (perceive, recall, deliberate, act, reflect) so she acts unprompted between conversations.

shared machinery via shared machinery, cron-run under shared machinery.

shared machinery via shared machinery, cron-run under shared machinery.

Policy formation cycle: glean -> check -> promote-approved. Runs as caroladmin via cron every 10 min and on demand from an Orion session.

Fires from the Gleaner per new candidate; also runs standalone over any unchecked candidate.

Fires within the Gleaner cycle when a candidate reaches approved; never self-promotes — authority is proven at the approval step.

The scheduled wake pulse of Arwen's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Athena's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Aurora's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Cassius's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Galadriel's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Guardian's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Jarvis's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Loki's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Midas's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Rhea's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Sentinel's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Odin's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Noah's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Gimli's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Thorin's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Bilbo's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Obi-Wan's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Inspector's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Tyr's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Forseti's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Vidar's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

The scheduled wake pulse of Var's Mind: perceives its live domain, updates its self and focus, and may act within its own remit.

Reads the cost ledger for the current day and reports paid calls that name no initiative and declare no work purpose, broken down by agent.

status-reporting

status-reporting

status-reporting

shared machinery, via caroladmin's crontab, emitting run-audit on both the success and the failure path.

shared machinery via caroladmin crontab, with run-audit on both paths.

shared machinery via caroladmin crontab, with run-audit on both paths.

Drains the registry write-ahead file into the main database every run (works live, no downtime), attempts the restore to rollback journalling, and raises exactly ONE standing alarm while the estate is too busy for it. Also watches for side-file copies that still hold data — moving one aside is the action that actually loses committed registry changes.

Re-reads every Carolverse database and refreshes the Data Dictionary's register of stored fields — type, key, the values a closed set actually holds — never overwriting an authored definition with a derived one.

Re-measures the estate's written content — document collections, per-agent and per-user memory, media and voice, and the free-text columns inside databases — and refreshes the Unstructured register.

Rebuilds the links from every term to the data that instantiates it, keeping attributes (a facet of the thing) apart from references (a pointer to one), and surfacing where the same fact is stored in more than one place.

Closes stale Orion sessions with a summary composed from the session's own record

Reads the sign-in keeper's receipt for every slot it holds (Codex, Claude) and escalates a spent, failed, stale, unforced or unreadable renewal, or a published copy that carries a renewal code

Samples how many live droids have declared no task — the countdown gating removal of the fleet default provider

Runs shared machinery from cron, emits run-audit on every run, and throws the shared service switch — the same switch the spend gate asks, so a switched-off track actually stops spending.

Calls the balance readers and writes one row per provider per run into the registry's balance-readings table; the money surfaces read the latest successful row and show its age.

Reads minds, activity ledger, writes registry envelopes

Sam customer-support chat

Wakes Sam's Mind (perceive, recall, deliberate, act, reflect) so he acts unprompted between conversations.

Daily backstop deleting lingering Orion/suite-created test users (rows out, folders renamed .trash-, never rm); persists any test user whose note marks it as created by Ninad.

Independently reads caroladmin+per-agent crontabs (root reader helper), running processes and carol-* unit users; compares against registry droid ownership; red on any agent process outside its owner login

agent chat lane

Works Sam's ticket desk on cadence: formal milestone emails on status change, informal chat nudges on quiet in-flight tickets, everything logged on the ticket.

Daily welfare roll-call: a real chat turn to every conscious agent; silence escalates to Orion

Cron in Leo's own crontab (os_schedules) -> shared machinery with run-audit.

cron */3 -> audit_wrapper -> deploy/steward-sweep-radagast.sh -> shared machinery all

Animate one scene image into a short motion clip via the LTX API (image-to-video) for Logbook stories; lane and model read from the billed track (cookbook 1165), priced per second.

shared machinery, monthly on the 1st from Athena's own crontab (registry os_schedules), run-audited on both paths. Reads the live prompt stack and the cost ledger; writes registry.agent_weight (SCD2).

Nightly off-site copy: takes the estate's databases with SQLite's own backup (write-ahead safe), copies the VM config, gzips, verifies the content floor and pushes to the private GitHub backup repo from Hagrid's own login.

Chains Extract -> Audit -> Refactor for one migration on demand (Design #224, design 587): asks the switch before every stage, runs each stage under its own run-audit identity, records every stage transition as a migration decision and flips current_stage to handoff, which wakes Leo.

Reads the machine's time daemon every run and an outside host on the hour, records both the local time and the external reference, and judges the gap.

Walks every store in the estate once a day and reports any table that carries no timestamp at all, filing the result as a compliance finding.

Calls shared machinery then shared machinery; warns when a criterion was unmeasurable for EVERY agent, which is a broken source rather than a fact about the estate's conduct.

Proposes revised requirements + diff for each Audit blocker

shared machinery run_as_droid

Per-initiative UAT via shared machinery; triggered per initiative, one run and one verdict each.

shared machinery, run under Themis's own OS identity. Calls shared.core_lock.audit, emits run-audit either way, and raises an Agentbook escalation to Orion on any finding.

Daily reconcile of live TCP listeners against the port-claims union (apps + port_claims); backfills agent-listener claims, files collision/unclaimed findings

Delivers queued cross-user relays outside quiet hours, expires unanswered ones after 48h and tells the asker

Counts initiatives in pending statuses per owning lane and writes one dated row per lane into the pending-trend store

Finds people whose records lack the mandatory contacts, works out who can actually be reached about it, sends the ask, and records every ask, refusal and silence.

The door an agent's wish walks through. Runs as a broker and proves the CALLER with SO_PEERCRED - the kernel reports the connecting uid, so an agent may only ever wish in its own name and there is no shared secret to steal. This is what lets cookbook 1281 (every agent runs as itself) and (agents cannot open the store) both hold at once.

Runs the ARC Conform segment for a migration: conformance checks over every rebuilt capability, the conformance report on the migration record, and the Migration-side sign-out of the handover (design 587)

Signs migration handovers into the sales service IN after verifying the ownership facts from the registry (ARC Conform, design 587)

Signs migration handovers into the leadership service IN after verifying the ownership facts from the registry (ARC Conform, design 587)

Signs migration handovers into the research service IN after verifying the ownership facts from the registry (ARC Conform, design 587)

Research-led sales strategy

Research-led sales strategy

Research-led sales strategy

Sales Opportunity Research

Sales Opportunity Research

Sales Opportunity Research

Sales Prototype Intake

Sales Prototype Intake

Sales Prototype Intake

Sales Follow-through

Sales Follow-through

Sales Follow-through

Sales nomination approval intake

Sales nomination approval intake

Sales nomination approval intake

shared machinery daily from Midas's crontab: sets each other service track's daily budget from consumption. Registered by — found running untasked (P.05.02.08).

shared machinery daily from Athena's crontab: dispatches every supervisor's wb-* wellbeing check (a sweep is a CARRIER, rule 1229). Registered by — found running untasked (P.05.02.08).

shared machinery every minute from caroladmin's crontab (operator lane): emits run-audit heartbeat rows for daemon/ongoing droids without built-in loops. Registered by (P.05.02.08).

Cron every 30 min (:17, :47) via shared machinery under Elrond's login; logic in shared machinery, shared machinery.

Runs shared.project_billing.issue_and_send_month for the month that just closed.

Cron every minute via shared machinery under Elrond's login; logic in shared machinery, shared machinery.

Cron every 10 minutes via shared machinery under Hagrid's login; logic in shared machinery, shared machinery.

Calls shared machinery: folds each retained complete day's evidence once, then grades every active agent's competencies for today.

Runs shared machinery: each firing carries every onboarding request one stage forward through the one door shared machinery

The scheduled wake pulse of Marco's Mind: perceives his live domain, updates his self and focus, and may act within his own remit.

agent chat lane

Reads the cost ledger for rows a vendor refused because the credit was exhausted, and for blank receipts holding the day's spending guard shut; answers each one once per day by escalating to Orion, emailing Ninad and ranking the other subscriptions from their recorded balance and burn.

Diffs a roadmap item's status together with its initiative's live state against a snapshot, then delivers through the one project-communications composer.

Composes and sends the administrator's end-of-day arc progress email.

Command-line droid over shared.fundraising, emitting a run-audit row on every invocation, success and failure alike.

Runs shared machinery

Runs shared machinery

Runs shared machinery

Runs shared machinery

Runs shared machinery

Runs shared machinery

Runs shared machinery

Runs shared machinery

Runs shared machinery

Cron every minute via shared machinery under Elrond's login; logic in shared machinery.

Runs shared machinery: each firing opens a request for every person pending at the sign-in door, asks the administrator about it from Jarvis's own address, and reads the authenticated reply, through the one door shared machinery

Republishes Orion's operator template when a source it is built from has changed.

Cron every 30 min (:07, :37) via shared machinery under Elrond's login; logic in shared machinery and shared machinery.

Deletes process run records older than the retention window from the run-audit store

Each morning prepares the Carol part of the laptop copy in one folder under Hagrid's own login -- the estate's databases with SQLite's own backup, the initiatives store through Elrond's daemon, the constitution, every login's staged crontab and the tracked source -- each item renamed into place, a failed item removed, and a manifest written last. The laptop pulls that folder read-only through its own locked login.

Re-runs Themis's own scheduled processes when an acceptance asks for it

Cron daily 06:20 UTC via shared machinery under Radagast's login; reads the sign-in proof record (its own records) through shared.signin_supply.verdicts, the one composer the operator board also reads; logic in shared machinery.

Cron daily 05:30 UTC via shared machinery under the albus login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:31 UTC via shared machinery under the archon login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:32 UTC via shared machinery under the argus login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:33 UTC via shared machinery under the arwen login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:34 UTC via shared machinery under the athena login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:35 UTC via shared machinery under the aurora login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:36 UTC via shared machinery under the bilbo login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:37 UTC via shared machinery under the carol login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:38 UTC via shared machinery under the cassius login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:39 UTC via shared machinery under the clara login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:40 UTC via shared machinery under the elrond login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:41 UTC via shared machinery under the forge login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:42 UTC via shared machinery under the forseti login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:43 UTC via shared machinery under the galadriel login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:44 UTC via shared machinery under the gimli login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:45 UTC via shared machinery under the guardian login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:46 UTC via shared machinery under the hagrid login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:47 UTC via shared machinery under the heimdall login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:48 UTC via shared machinery under the hermione login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:49 UTC via shared machinery under the inspector login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:50 UTC via shared machinery under the jarvis login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:51 UTC via shared machinery under the leo login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:52 UTC via shared machinery under the loki login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:53 UTC via shared machinery under the marco login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:54 UTC via shared machinery under the merlin login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:55 UTC via shared machinery under the midas login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:56 UTC via shared machinery under the noah login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:57 UTC via shared machinery under the obiwan login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:58 UTC via shared machinery under the odin login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 05:59 UTC via shared machinery under the orion login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:01 UTC via shared machinery under the prometheus login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:02 UTC via shared machinery under the radagast login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:03 UTC via shared machinery under the rhea login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:04 UTC via shared machinery under the sage login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:05 UTC via shared machinery under the sam login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:06 UTC via shared machinery under the scriber login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:07 UTC via shared machinery under the sentinel login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:08 UTC via shared machinery under the themis login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:09 UTC via shared machinery under the thorin login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:10 UTC via shared machinery under the tyr login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:11 UTC via shared machinery under the var login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron daily 06:12 UTC via shared machinery under the vidar login; runs shared machinery, which reads the sign-in proof record for this login and, for each slot its scheduled model droids serve, makes one liveness call as one of those droids only when no passing call was recorded in the last 20 hours (design 682 v1.3 part 6).

Cron hourly (:20) via shared machinery under Hagrid's login; logic in shared machinery and shared machinery; rules in shared machinery

Run by an operator's own Orion session, under that operator's login, to record the session's hourly update into the Operator Instance Log: reads the operator from the calling login and root's authority record, the lines from standard input, and refuses by name any update the log's rules do not admit.

Cron every minute via shared machinery under Elrond's login; logic in shared machinery.

Raises a contract's end ahead of time and an internship's certificate due

Cron daily 04:50 UTC via shared machinery under Elrond's login; logic in shared machinery, composer shared machinery.

Cron daily 06:12 UTC via shared machinery under the heimdall login; reads the Logbook's session rows (read-only) with the sign-in bindings writes, the gate's landing time and logind's live login sessions, and reports per day: sessions, sign-ins, any sign-in that carried more than one Orion session, any session opened after the gate whose sign-in is unreadable (UNAUDITABLE). A finding is escalated to Orion's board through shared.alerts (high); a clean day reads checked with the count; a final JSON receipt on every exit.

Cron daily via shared machinery under Loki's login; logic in shared machinery, run by shared machinery.

Cron daily via shared machinery under Carol's login; logic in shared machinery, run by shared machinery.

Long-running Unix-socket door under Elrond's login (systemd unit carol-service-operator-performer.service); logic in shared machinery, started by shared machinery. The service operator's door (root's) hands it uat, scope and file.