The loop graph

The control loops this repository declares, projected from the loop registry at .aof/loops. Regenerate with aof work loops document --write after any change to a loop record.

Health

• Declared records: 17
• Declared edges: 23
• Findings: 0 error, 33 warning

The diagram below draws every endpoint a record points at, including endpoints no record declares; the record count above counts declared records only.

The graph

flowchart LR
  actor_operator(["actor:operator · Human operator"])
  actor_product_owner(["actor:product-owner · Product owner"])
  anchor_rubric_process_exit(("anchor:rubric-process-exit · Rubric process exit"))
  anchor_run_lifecycle_policy(("anchor:run-lifecycle-policy · Run lifecycle policy"))
  anchor_run_liveness(("anchor:run-liveness · Run liveness observation"))
  arbiter_speed_thoroughness_autonomy{"arbiter:speed-thoroughness-autonomy · Speed versus thoroughness versus autonomy"}
  auditor_instrument_audit>"auditor:instrument-audit · The instruments are audited by something none of them supervises"]
  config_work_autonomous_maxAttempts[/"config:work.autonomous.maxAttempts"/]
  config_work_loop_buildNoProgressRounds[/"config:work.loop.buildNoProgressRounds"/]
  config_work_loop_reviewRounds[/"config:work.loop.reviewRounds"/]
  loop_autonomous_cascade["loop:autonomous-cascade · Advance a work range to done"]
  loop_build_to_green["loop:build-to-green · Build executable work to green"]
  loop_mesh_assignment_reclaim["loop:mesh-assignment-reclaim · Reclaim assignments only after dual staleness"]
  loop_retrospective_memory_ingest["loop:retrospective-memory-ingest · Capture milestone lessons into memory"]
  loop_review_fix_rereview["loop:review-fix-rereview · Review, fix, and re-review"]
  loop_run_resilience["loop:run-resilience · Keep runs within their lifecycle policy"]
  loop_verify_triage_accept["loop:verify-triage-accept · Verify, triage, and accept an item"]
  watcher_autonomous_cascade_watcher{{"watcher:autonomous-cascade-watcher · Autonomous cascade watched against intervention rate"}}
  watcher_build_to_green_watcher{{"watcher:build-to-green-watcher · Build-to-green watched against contract shrinking"}}
  watcher_review_fix_rereview_watcher{{"watcher:review-fix-rereview-watcher · Review-fix-rereview watched against finding escape"}}
  actor_operator -->|target-setting| arbiter_speed_thoroughness_autonomy
  actor_operator -->|target-setting| loop_autonomous_cascade
  actor_operator -->|target-setting| loop_mesh_assignment_reclaim
  actor_operator -->|target-setting| loop_retrospective_memory_ingest
  actor_product_owner -->|target-setting| loop_verify_triage_accept
  anchor_rubric_process_exit -->|data-feed| loop_build_to_green
  anchor_run_lifecycle_policy -->|target-setting| loop_run_resilience
  anchor_run_liveness -->|data-feed| loop_mesh_assignment_reclaim
  anchor_run_liveness -->|data-feed| loop_run_resilience
  arbiter_speed_thoroughness_autonomy -->|parameter-tuning| config_work_autonomous_maxAttempts
  arbiter_speed_thoroughness_autonomy -->|parameter-tuning| config_work_loop_buildNoProgressRounds
  arbiter_speed_thoroughness_autonomy -->|parameter-tuning| config_work_loop_reviewRounds
  arbiter_speed_thoroughness_autonomy -->|veto| loop_autonomous_cascade
  arbiter_speed_thoroughness_autonomy -->|veto| loop_build_to_green
  arbiter_speed_thoroughness_autonomy -->|veto| loop_review_fix_rereview
  arbiter_speed_thoroughness_autonomy -->|veto| loop_verify_triage_accept
  auditor_instrument_audit -->|reporting| actor_operator
  auditor_instrument_audit -->|reporting| actor_product_owner
  loop_autonomous_cascade -->|target-setting| loop_build_to_green
  loop_autonomous_cascade -->|target-setting| loop_review_fix_rereview
  watcher_autonomous_cascade_watcher -->|monitoring| loop_autonomous_cascade
  watcher_build_to_green_watcher -->|monitoring| loop_build_to_green
  watcher_review_fix_rereview_watcher -->|monitoring| loop_review_fix_rereview

The records

actor:operator – Human operator

• kind: actor
• ground: exogenous

actor:product-owner – Product owner

• kind: actor

anchor:rubric-process-exit – Rubric process exit

• kind: anchor
• ground: process-exit
• observes: module:src/commands/grade.mjs#reportObservation

anchor:run-lifecycle-policy – Run lifecycle policy

• kind: anchor
• ground: frozen-rule
• observes: module:src/run-store.mjs#isLegalTransition

anchor:run-liveness – Run liveness observation

• kind: anchor
• ground: live-soak
• observes: module:src/run-store.mjs#isStale

arbiter:speed-thoroughness-autonomy – Speed versus thoroughness versus autonomy

• kind: arbiter
• resolves: how much of the same agent’s effort each loop may spend
• priority: loop:verify-triage-accept, loop:review-fix-rereview, loop:build-to-green, loop:autonomous-cascade
• dwell: cycles:2

auditor:instrument-audit – The instruments are audited by something none of them supervises

• kind: auditor
• audits: module:scripts/test.mjs#tests, module:src/work-audit/census.mjs#runCensus, module:src/work-audit/evidence.mjs#runEvidence, module:src/work-audit/spawn.mjs#runBounded, module:src/work/doctor-controls.mjs#fitnessDeclarations, module:src/work/loops-checks.mjs#buildGroundednessReport, command:work:loops-validate, watcher:autonomous-cascade-watcher, watcher:build-to-green-watcher, watcher:review-fix-rereview-watcher, anchor:rubric-process-exit, anchor:run-lifecycle-policy, anchor:run-liveness
• measurement: command:work:audit
• cadence: event:per-milestone
• escalation: actor:operator

loop:autonomous-cascade – Advance a work range to done

• kind: loop
• controlled: items reaching done over a work range
• reference: command:work:next
• measurement: command:work:next
• actuator: prose:src/bundle/agents/aof-product-owner.md, prose:src/bundle/agents/aof-developer.md, prose:src/bundle/agents/aof-qa.md
• cadence: event:per-item
• ceiling: config:work.autonomous.maxAttempts
• owner: unknown
• optimizing: true
• layer: management

loop:build-to-green – Build executable work to green

• kind: loop
• controlled: executable scenarios and fitness functions green
• reference: prose:src/bundle/commands/continue.md
• measurement: prose:src/bundle/commands/continue.md
• actuator: prose:src/bundle/agents/aof-developer.md
• cadence: event:per-phase
• ceiling: config:work.loop.buildNoProgressRounds
• owner: unknown
• optimizing: true
• layer: operational

loop:mesh-assignment-reclaim – Reclaim assignments only after dual staleness

• kind: loop
• controlled: module:src/mesh/assignment-reclaim.mjs#reclaimStaleAssignments
• reference: module:src/mesh/assignment-reclaim.mjs#dualStalenessDecision, module:src/mesh/presence.mjs#isNodeStale, module:src/run-store.mjs#isStale
• measurement: module:src/mesh/assignment-reclaim.mjs#dualStalenessDecision, module:src/mesh/presence.mjs#isNodeStale, module:src/run-store.mjs#isStale
• actuator: module:src/effects/assignment-transitions.mjs#transitionAssignmentState, module:src/effects/run-transitions.mjs#transitionRunReclaimed
• cadence: periodic:15s
• ceiling: none
• owner: unknown
• optimizing: false
• layer: operational

loop:retrospective-memory-ingest – Capture milestone lessons into memory

• kind: loop
• controlled: milestone lessons made recallable
• reference: prose:src/bundle/commands/retrospective.md
• measurement: prose:src/bundle/commands/retrospective.md
• actuator: module:src/work/memory.mjs#runMemory
• cadence: event:per-milestone
• ceiling: none
• owner: unknown
• optimizing: false
• layer: governance

loop:review-fix-rereview – Review, fix, and re-review

• kind: loop
• controlled: open review findings
• reference: prose:src/bundle/commands/continue.md
• measurement: prose:src/bundle/commands/continue.md
• actuator: prose:src/bundle/agents/aof-developer.md
• cadence: event:per-phase
• ceiling: config:work.loop.reviewRounds
• owner: unknown
• optimizing: true
• layer: operational

loop:run-resilience – Keep runs within their lifecycle policy

• kind: loop
• controlled: module:src/run-store.mjs#readRuns
• reference: module:src/run-store.mjs#isLegalTransition, module:src/run-store.mjs#isRetryable
• measurement: module:src/run-store.mjs#isStale, module:src/run-store.mjs#retryReadiness
• actuator: command:work:run-start, command:work:run-retry, command:work:run-complete
• cadence: event:per-run-start
• ceiling: config:work.autonomous.maxAttempts, module:src/run-store.mjs#shouldRetry
• owner: unknown
• optimizing: false
• layer: operational

loop:verify-triage-accept – Verify, triage, and accept an item

• kind: loop
• controlled: findings triaged and item accepted
• reference: prose:src/bundle/commands/verify.md
• measurement: prose:src/bundle/commands/verify.md
• actuator: prose:src/bundle/agents/aof-developer.md, prose:src/bundle/agents/aof-qa.md, prose:src/bundle/agents/aof-product-owner.md
• cadence: event:per-item
• ceiling: none
• owner: actor:product-owner
• optimizing: false
• layer: management

watcher:autonomous-cascade-watcher – Autonomous cascade watched against intervention rate

• kind: watcher
• counter: how often a run needed a retry or a hand
• determinism: counter
• measurement: command:work:counters

watcher:build-to-green-watcher – Build-to-green watched against contract shrinking

• kind: watcher
• counter: whether the acceptance criteria got smaller
• determinism: counter
• measurement: command:work:ratchet

watcher:review-fix-rereview-watcher – Review-fix-rereview watched against finding escape

• kind: watcher
• counter: findings raised after the item was accepted
• determinism: counter
• measurement: command:work:counters