Promote dev → main: agent card + fleet-health filter + ceremony fixes (v0.7.22 candidate, re-cut)

[mabry1985]

Re-cut of #463 after back-merge #464 cleared the merge-base drift. Full PR list (17 fixes since v0.7.21):

PR	Description
#427	wire alert.* executors + startup validator
#428	stop world.state.# leak from ceremony snapshots
#429	rollcall skill source
#431	wire ceremony skill executors
#432	wire 5 pr-remediator skills
#433	register protomaker A2A agent
#434	drop overscoped subscribesTo from protomaker
#435	rename web_search → searxng_search in ava.yaml
#436	disable fleet_agent_stuck loop (mitigation)
#451	disable 4 more action.* loops (mitigation)
#452	per-action cooldown infrastructure (closes #437)
#454	route 4 ceremonies, disable 3 broken
#455	pre-dispatch target registry guard (closes #444)
#456	honor enabled:false on initial ceremony load (closes #453)
#457	wrap subscribe-spy push callbacks to return void (CI green)
#460	filter outcomes from synthetic actors (closes #459)
#462	agent card advertises canonical A2A endpoint (closes #461)

All landed and verified live on :dev for 24+ hours. Discord ops silent for the duration (cooldown + registry guard holding).

HITL-gated per protocol — operator approval required. Do not auto-merge.

Note: pr-remediator's auto-close-on-decomposable misfired on the original #463 (filed as #465). Code fix not yet merged, so this re-cut is at risk of being auto-closed AGAIN if a new conflict surfaces while it's open. Mitigation: assuming the back-merge cleaned everything, this should be conflict-free at open time.

Summary by CodeRabbit

Release Notes

• New Features

  • Action cooldown support: Actions can now define per-action throttling windows to prevent excessive repeated execution.
  • New environment variables for A2A agent discovery: WORKSTACEAN_PUBLIC_BASE_URL and WORKSTACEAN_INTERNAL_HOST control externally-advertised agent endpoint URLs.
  • Agent targeting validation: Dispatch failures now occur when action targets reference non-existent agents.
  • Fleet health insights: System actors now reported separately from registered agents in health snapshots.

• Bug Fixes

  • Ceremony state updates no longer leak into world-state message topics, preventing false goal evaluations.
  • Malformed world-state payloads are now silently ignored; goals no longer report spurious violations.
  • Fixed selector validation to detect broken goal selectors at startup.

• Documentation

  • Updated environment variable examples and A2A agent-card resolution logic.
  • Ceremony disabling: Several ceremonies with missing executor skills are now disabled by default.

• Configuration

  • Added action cooldown values across all GOAP actions.
  • Refined ceremony skill targets to specific agents.
  • Updated agent tools configuration.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR implements multi-layer fixes for duplicate action dispatches, unregistered agent routing failures, and ceremony state isolation. It introduces action-level cooldown deduplication in ActionDispatcherPlugin, adds pre-dispatch registry validation to block phantom targets, separates ceremony state publishing from world state, implements three new skill executor plugins, and fixes ceremony disabled-state handling on initial load.

Changes

Cohort / File(s)	Summary
A2A Agent Card URL Resolution src/api/agent-card.ts, src/api/__tests__/agent-card.test.ts, src/config/env.ts, .env.dist, README.md, docs/reference/env-vars.md, docs/reference/http-api.md	Replaced WORKSTACEAN_BASE_URL-based card url derivation with new env vars: WORKSTACEAN_PUBLIC_BASE_URL (preferred) and WORKSTACEAN_INTERNAL_HOST (fallback, default workstacean). Added resolveA2aUrl() helper and additionalInterfaces[] to agent card. Updated docs and examples.
Action Cooldown Deduplication src/planner/types/action.ts, src/schemas/yaml-schemas.ts, src/plugins/action-dispatcher-plugin.ts, src/plugins/action-dispatcher-plugin.test.ts, workspace/actions.yaml	Added optional meta.cooldownMs field to action metadata. Implemented per-action cooldown throttling in ActionDispatcherPlugin using lastDispatchedAt state and _admitOrCooldown() gate; repeated dispatches within window are dropped and telemetry-bumped (cooldown_dropped). Added 30+ actions with default cooldowns (alert 15min, ceremony 30min, pr-remediator 5min).
Pre-Dispatch Registry Guard & Executor Validation src/plugins/action-dispatcher-plugin.ts, src/plugins/action-dispatcher-plugin.test.ts, src/planner/validate-action-executors.ts, src/planner/__tests__/validate-action-executors.test.ts, src/index.ts	Added executorRegistry constructor parameter to ActionDispatcherPlugin; new _admitOrTargetUnresolved() gate checks meta.agentId against registered executors before dispatch (blocks with target_unresolved telemetry). New validation module exports findUnwiredActions(), UnwiredActionsError, validateActionExecutors() for startup cross-check; runs after plugin installation when WORKSTACEAN_STRICT_WIRING=1.
Three Skill Executor Plugins src/plugins/alert-skill-executor-plugin.ts, src/plugins/alert-skill-executor-plugin.test.ts, src/plugins/ceremony-skill-executor-plugin.ts, src/plugins/ceremony-skill-executor-plugin.test.ts, src/plugins/pr-remediator-skill-executor-plugin.ts, src/plugins/pr-remediator-skill-executor-plugin.test.ts	Added AlertSkillExecutorPlugin (registers alert.* skills, publishes to message.outbound.discord.alert), CeremonySkillExecutorPlugin (registers ceremony.* actions, publishes to ceremony.<id>.execute), PrRemediatorSkillExecutorPlugin (registers action.pr_* skills, publishes to pr-remediator topics). Each plugin registers FunctionExecutor instances with fixed priority and implements fire-and-forget semantics.
Ceremony Plugin Fixes & State Separation src/plugins/CeremonyPlugin.ts, src/plugins/__tests__/CeremonyPlugin.test.ts, src/world/extensions/CeremonyStateExtension.ts, src/world/extensions/__tests__/CeremonyStateExtension.test.ts, docs/reference/ceremony-plugin.md, src/loaders/ceremonyYamlLoader.ts	Fixed initial-load handling of enabled: false ceremonies (skip registration, timers, state-extension). Changed CeremonyStateExtension to publish snapshots to ceremony.state.snapshot instead of world.state.snapshot with explicit ban on world.state.# namespace. Added hot-reload logic to detect disabled ceremonies and unregister them (cancel timers, remove extensions).
Fleet Health Synthetic Actor Filtering src/plugins/agent-fleet-health-plugin.ts, src/plugins/agent-fleet-health-plugin.test.ts	Added executorRegistry parameter to AgentFleetHealthPlugin; new systemActors[] field in FleetHealthSnapshot and separate systemActorWindows map. Outcomes from non-registered actors route to systemActors instead of agents[], preventing synthetic names (pr-remediator, auto-triage-sweep) from inflating agent metrics. _isRegisteredAgent() check gates routing; legacy behavior preserved when registry absent.
Goal Evaluator Payload Validation src/plugins/goal_evaluator_plugin.ts, __tests__/goal_evaluator_plugin.test.ts	Added _looksLikeWorldState() guard to validate incoming world.state.# payloads before evaluation; non-matching payloads ignored with warning. Implemented validatorRan flag to run selector resolution validation exactly once after first evaluated snapshot. Added helper methods _describePayload() and _validateLoadedGoalSelectors() for logging and selector path checks (issue #424).
HITL Policy Support lib/plugins/pr-remediator.ts, src/pr-remediator.test.ts	Added _extractHitlPolicy(msg) to read msg.payload.meta.hitlPolicy (supports ttlMs and onTimeout). Extended _emitHitlApproval(hitlPolicy?) to accept optional policy; when provided, overrides request's expiresAt duration and conditionally adds onTimeout into emitted HITLRequest.options fields.
Ceremony & Agent Configuration workspace/ceremonies/agent-health.yaml, workspace/ceremonies/board.cleanup.yaml, workspace/ceremonies/board.health.yaml, workspace/ceremonies/board.retro.yaml, workspace/ceremonies/daily-standup.yaml, workspace/ceremonies/health-check.yaml, workspace/ceremonies/service-health.yaml, workspace/agents.yaml	Disabled 4 ceremonies with no advertised skills (agent-health, board.retro, service-health); updated targets for 4 ceremonies to route skills to correct agents (board.cleanup, board.health, daily-standup, health-check → quinn or protomaker). Added protomaker agent entry with A2A endpoint and apiKey auth.
Documentation & Configuration docs/reference/env-vars.md, docs/reference/http-api.md, docs/reference/ceremony-plugin.md, README.md, .env.dist, .claude/commands/rollcall.md	Updated docs to reflect A2A URL resolution logic, new env vars, and ceremony state separation. Changed rollcall smoke test path and scope documentation.
Telemetry & Tests src/telemetry/telemetry-service.ts, src/planner/__tests__/end-to-end-loop.test.ts, __tests__/goal_evaluator_plugin.test.ts	Added "cooldown_dropped" to ActionEvent type and ACTION_EVENTS constant. Added end-to-end test for cooldown suppression (100 violations → 1 executor dispatch). Expanded goal evaluator tests for payload shape rejection and selector validation.

Sequence Diagram(s)

sequenceDiagram
    participant Planner as GOAP Planner
    participant Dispatcher as ActionDispatcher
    participant Registry as ExecutorRegistry
    participant Executor as FunctionExecutor
    participant Bus as EventBus

    Planner->>Dispatcher: handleDispatch(action)
    
    rect rgba(255, 165, 0, 0.5)
    Note over Dispatcher: Cooldown Gate
    Dispatcher->>Dispatcher: _admitOrCooldown(action)
    alt Within cooldown window
        Dispatcher->>Dispatcher: log warning, bump telemetry
        Dispatcher-->>Planner: (dropped)
    end
    end
    
    rect rgba(100, 150, 255, 0.5)
    Note over Dispatcher: Target Registry Check
    Dispatcher->>Registry: resolve(skillHint, targets)
    Registry-->>Dispatcher: executor or null
    alt Target unregistered
        Dispatcher->>Dispatcher: log warning, bump target_unresolved
        Dispatcher-->>Planner: (dropped)
    end
    end
    
    rect rgba(144, 238, 144, 0.5)
    Note over Executor: Executor Dispatch
    Dispatcher->>Executor: _execute(skillRequest)
    Executor->>Bus: publish(message)
    Bus-->>Executor: ack
    Executor-->>Dispatcher: SkillResult
    end
    
    Dispatcher->>Dispatcher: record lastDispatchedAt
    Dispatcher-->>Planner: return

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~75 minutes

---

Possibly related issues

• fix(goap): action.fleet_incident_response and alert.fleet_agent_stuck need dedup/cooldown — fired duplicate INC-003 through INC-009 in 30s #437 (action cooldown): Implemented per-action cooldownMs deduplication in ActionDispatcherPlugin to prevent duplicate incident filing; directly resolves the suggested "Action-level cooldown" layer.
• fix(goap): pre-dispatch registry check missing — dispatches to unregistered agents (auto-triage-sweep, user) cause 100% failure rate #444 (pre-dispatch registry guard): Implemented _admitOrTargetUnresolved() check in ActionDispatcherPlugin to validate meta.agentId against ExecutorRegistry before dispatch; directly resolves the suggested "pre-dispatch registry check missing" issue.
• fix(ceremony-loader): enabled: false not honored on initial load — disabled ceremonies still fire on cron #453 (ceremony enabled: false on initial load): Fixed CeremonyPlugin.install() to filter disabled ceremonies before scheduling cron timers, and added hot-reload logic to unregister disabled ceremonies that transition from enabled→disabled state.
• fix(fleet-health): don't attribute outcomes to synthetic/plugin actors — pr-remediator, auto-triage-sweep show as 'stuck agents' #459 (synthetic actor filtering): Implemented separate systemActors[] aggregation in AgentFleetHealthPlugin using ExecutorRegistry whitelist; outcomes from non-registered actors (pr-remediator, auto-triage-sweep) now route to systemActors instead of polluting agents[] metrics.
• fix(a2a): agent card advertises wrong url (ava:8081 — the dashboard) instead of the A2A JSON-RPC endpoint (workstacean:3000/a2a) #461 (agent card URL mismatch): Replaced hardcoded WORKSTACEAN_BASE_URL derivation with WORKSTACEAN_PUBLIC_BASE_URL (preferred) and WORKSTACEAN_INTERNAL_HOST (fallback) for correct A2A endpoint discovery.

---

Possibly related PRs

• fix(ceremonies): route 4 ceremonies to right skill owner, disable 3 broken #454 (ceremony routing & disabling): Overlaps in ceremony target updates (board.cleanup→quinn, board.health→protomaker, daily-standup/health-check→quinn) and ceremony disables (agent-health, board.retro, service-health); this PR consolidates those fixes alongside the architectural cooldown/registry/separation changes.

---

Poem

🐰 A cooldown thumps the hasty planner's heart,
Registry whispers "agent, thou dost exist,"
No more phantom dispatches tear apart—
Ceremonies bow to their enabled list.
The fleet breathes clean; the card points true. ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 54.55% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title comprehensively summarizes the main changes: promoting dev to main with agent card fixes, fleet-health filtering, and ceremony fixes for v0.7.22.
Linked Issues check	✅ Passed	The PR implements all coding requirements from linked issues #437, #444, #453, #459, and #461 with per-action cooldowns, pre-dispatch registry guards, disabled ceremony handling, synthetic actor filtering, and agent card URL fixes.
Out of Scope Changes check	✅ Passed	All code changes directly address the linked issue requirements: cooldown infrastructure, executor registry guards, ceremony enabling logic, fleet-health filtering, and agent card URL resolution. Tool rename (searxng_search) and ceremony routing/disabling align with #444 audit guidance.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch dev

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 6

🧹 Nitpick comments (6)

.claude/commands/rollcall.md (1)

1-1: Good terminology refinement; consider hyphenation.

The updated description "(in-process DeepAgents, A2A external services)" is more precise and aligns well with the codebase terminology shown in the relevant code snippets.

Minor style note: As a compound adjective modifying "smoke test," "roll call" could be hyphenated as "roll-call."

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In @.claude/commands/rollcall.md at line 1, Update the sentence "Run the agent
roll call smoke test. Reports status of all agents (in-process DeepAgents, A2A
external services), AI infrastructure, media stack, and monitoring services." to
hyphenate the compound adjective by changing "roll call" to "roll-call" (i.e.,
"Run the agent roll-call smoke test...") while preserving the parenthetical
"(in-process DeepAgents, A2A external services)" and the rest of the
description.

workspace/ceremonies/health-check.yaml (1)

5-12: Routing fix is correct; ceremony semantics are still misleading.

The target change to quinn is solid, but this config still executes board_audit under a service-health identity. Consider renaming/reframing this ceremony (or switching to an actual health-report skill) to avoid operator confusion.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@workspace/ceremonies/health-check.yaml` around lines 5 - 12, Summary: The
ceremony is routed correctly to quinn but its identity still reads as a
service-health check while running the board_audit skill, which is misleading.
Fix: change the ceremony identity or name and/or replace the skill to match
intent so operators aren't confused — either rename this ceremony from a
"service-health" framing to something that reflects a board audit, or swap the
skill to an actual health-report skill; specifically update the YAML entries
referencing the skill value "board_audit" and any metadata/label that implies
"service-health" so the ceremony name/skill and the identity are consistent with
each other and with the target "quinn".

__tests__/goal_evaluator_plugin.test.ts (1)

356-367: Capture console.error before install to make the startup-validator test robust.

If validator logging happens during install(), the current ordering can miss the signal and create a false-negative test.

Proposed test ordering tweak

-    plugin = new GoalEvaluatorPlugin({ workspaceDir: join(TMP_DIR, "workspace") });
-    plugin.install(bus);
-
     const errors: string[] = [];
     const origError = console.error;
     console.error = (...args: unknown[]) => {
       errors.push(args.map(a => (typeof a === "string" ? a : JSON.stringify(a))).join(" "));
     };
 
     try {
+      plugin = new GoalEvaluatorPlugin({ workspaceDir: join(TMP_DIR, "workspace") });
+      plugin.install(bus);
       bus.publish("world.state.updated", {
         id: "ws-2",
         correlationId: "ws-2",

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@__tests__/goal_evaluator_plugin.test.ts` around lines 356 - 367, The test
captures console.error too late: move the console.error override so it runs
before calling plugin.install(...) to ensure any validator logging during
install is caught; specifically, in the test that constructs GoalEvaluatorPlugin
and calls plugin.install(bus), set up the errors array and override
console.error prior to invoking plugin.install, then restore console.error after
bus.publish(...) completes, referencing GoalEvaluatorPlugin, install,
console.error and bus.publish to locate the affected lines.

src/plugins/CeremonyPlugin.ts (1)

538-547: Consider clearing ceremony history when unregistering.

When a ceremony is disabled via hot-reload, unregisterCeremony removes it from ceremonies and status maps but — per the context snippet from CeremonyStateExtension.ts:48-55 — does not clear state.history or state.lastRun. Old execution history may persist in snapshots, potentially affecting monitoring/reporting systems.

If stale history is acceptable (e.g., for audit trails), this is fine. Otherwise, consider extending unregisterCeremony to also clear history entries for the ceremony ID.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/plugins/CeremonyPlugin.ts` around lines 538 - 547, The unregisterCeremony
path currently removes entries from the ceremonies and status maps but leaves
per-ceremony execution data in CeremonyStateExtension (state.history and
state.lastRun), so update the unregisterCeremony method to also clear any stored
execution history and lastRun for the given ceremony id (e.g., remove
state.history[id] and state.lastRun[id] or call the appropriate
CeremonyStateExtension clear method) to avoid stale history persisting after a
hot-disable; ensure you reference and clear the same keys used by
CeremonyStateExtension.ts (state.history and state.lastRun) when unregistering.

src/plugins/__tests__/CeremonyPlugin.test.ts (1)

113-160: Consider increasing timing margins for CI stability.

The 5ms and 10ms delays may be insufficient on slow CI runners. Consider using slightly larger margins (e.g., 20-50ms) or a polling-based assertion pattern to reduce flakiness.

💡 Example polling pattern

// Instead of fixed delays:
const waitFor = async (fn: () => boolean, timeoutMs = 100) => {
  const start = Date.now();
  while (!fn() && Date.now() - start < timeoutMs) {
    await new Promise(r => setTimeout(r, 5));
  }
};
await waitFor(() => !plugin.getCeremonies().map(c => c.id).includes("board.flipping"));

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/plugins/__tests__/CeremonyPlugin.test.ts` around lines 113 - 160, The
test "CeremonyPlugin loader: hot-reload flips enabled→disabled (cancels timer,
removes from registry)" is flaky due to very small fixed delays (setTimeout 5ms
and 10ms); either replace those sleeps with a small polling helper (e.g.,
waitFor that repeatedly checks plugin.getCeremonies() or
timers.has("board.flipping") until the condition is met) and await it after
calling pluginAny._checkForChanges, or simply increase the delays to a safer
margin (e.g., 20–50ms). Update references in the test to use the polling helper
or larger timeouts around the pluginAny._checkForChanges calls that follow
writeCeremony so CI stability improves.

src/planner/__tests__/validate-action-executors.test.ts (1)

148-156: Consider using a more explicit type cast.

The bus as never cast works but is unconventional. Using as unknown as EventBus would make the intent clearer to future readers.

♻️ Suggested change

-    validateActionExecutors(actions, executors, { bus: bus as never });
+    validateActionExecutors(actions, executors, { bus: bus as unknown as EventBus });

Add the import at the top:

import type { EventBus } from "../../../lib/types.ts";

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/planner/__tests__/validate-action-executors.test.ts` around lines 148 -
156, The test uses an unconventional cast "bus as never" when calling
validateActionExecutors; instead import the EventBus type and cast the mock bus
more explicitly (e.g., bus as unknown as EventBus) so intent is clear — update
the test to add the import for EventBus and replace the "bus as never" cast with
"as unknown as EventBus", referencing the local bus variable and the
validateActionExecutors(...) call to locate where to change.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In @.claude/commands/rollcall.md:
- Line 5: Replace the hard-coded absolute path in the rollcall command with a
repo-relative invocation so the script is portable; update the line that
currently runs `bash /home/josh/dev/protoWorkstacean/scripts/agent-rollcall.sh`
in .claude/commands/rollcall.md to call the repo-local script (e.g.,
`scripts/agent-rollcall.sh` or `./scripts/agent-rollcall.sh`) or use a repo-root
resolution (e.g., via git rev-parse --show-toplevel) so the command works from
the repository regardless of the developer's home directory or CI environment.

In `@lib/plugins/pr-remediator.ts`:
- Around line 1201-1214: The extracted hitl policy currently allows ttlMs values
like NaN or Infinity which later cause RangeError in _emitHitlApproval; update
_extractHitlPolicy to validate ttlMs by checking typeof p.ttlMs === "number" &&
Number.isFinite(p.ttlMs) && p.ttlMs >= 0 (optionally Math.floor if you want an
integer) before assigning out.ttlMs so only finite, non-negative numbers are
propagated to _emitHitlApproval.

In `@README.md`:
- Line 65: Add a dedicated environment variable entry for
WORKSTACEAN_INTERNAL_HOST in the env-table: describe it as the internal host
used for docker-network fallback, note its default value (workstacean) and how
it interacts with WORKSTACEAN_HTTP_PORT and WORKSTACEAN_PUBLIC_BASE_URL (i.e.,
used in the fallback
`http://${WORKSTACEAN_INTERNAL_HOST:-workstacean}:${WORKSTACEAN_HTTP_PORT}/a2a`),
so readers can easily find and override the internal host behavior.

In `@src/index.ts`:
- Around line 511-526: The wiring validation currently runs too early; move the
validateActionExecutors call (the block that imports
"./planner/validate-action-executors.js" and calls
validateActionExecutors(actionRegistry, executorRegistry, { bus, throwOnUnwired:
process.env.WORKSTACEAN_STRICT_WIRING === "1" })) so it executes after
loadWorkspacePlugins() completes, and also invoke the same
validateActionExecutors call from the config reload path that handles
TOPICS.CONFIG_RELOAD (so actions reloaded at runtime get validated against
executorRegistry). Ensure you reuse the same actionRegistry and executorRegistry
instances and preserve the existing logging/throwOnUnwired behavior.
- Around line 316-329: The pr-remediator skill executor is unconditionally
installed (name: "pr-remediator-skill-executor", factory: returns new
PrRemediatorSkillExecutorPlugin(executorRegistry)) causing actions to be routed
when the actual pr-remediator plugin is absent; change the registrar's condition
from () => true to the same gating predicate used by the real pr-remediator
plugin (e.g., reuse the GitHub-credentials check or the plugin presence check),
or make the condition consult the plugin registry
(pluginManager.has("pr-remediator") / similar) so the
PrRemediatorSkillExecutorPlugin is only registered when the real pr-remediator
plugin will be installed.

In `@workspace/agents.yaml`:
- Around line 90-95: The protomaker agent registration references AVA_API_KEY
but that env var is not documented or provisioned; add AVA_API_KEY to the
environment template and onboarding/provisioning docs/scripts so fresh or local
deployments export a placeholder or real key before registering the agent.
Update the onboarding guide and any env files (the env template used for local
dev / CI) to include AVA_API_KEY with a short note about its purpose and where
to obtain/set it, and ensure any deployment/provision scripts (used during
bootstrap) populate or validate AVA_API_KEY before the protomaker agent is
registered.

---

Nitpick comments:
In `@__tests__/goal_evaluator_plugin.test.ts`:
- Around line 356-367: The test captures console.error too late: move the
console.error override so it runs before calling plugin.install(...) to ensure
any validator logging during install is caught; specifically, in the test that
constructs GoalEvaluatorPlugin and calls plugin.install(bus), set up the errors
array and override console.error prior to invoking plugin.install, then restore
console.error after bus.publish(...) completes, referencing GoalEvaluatorPlugin,
install, console.error and bus.publish to locate the affected lines.

In @.claude/commands/rollcall.md:
- Line 1: Update the sentence "Run the agent roll call smoke test. Reports
status of all agents (in-process DeepAgents, A2A external services), AI
infrastructure, media stack, and monitoring services." to hyphenate the compound
adjective by changing "roll call" to "roll-call" (i.e., "Run the agent roll-call
smoke test...") while preserving the parenthetical "(in-process DeepAgents, A2A
external services)" and the rest of the description.

In `@src/planner/__tests__/validate-action-executors.test.ts`:
- Around line 148-156: The test uses an unconventional cast "bus as never" when
calling validateActionExecutors; instead import the EventBus type and cast the
mock bus more explicitly (e.g., bus as unknown as EventBus) so intent is clear —
update the test to add the import for EventBus and replace the "bus as never"
cast with "as unknown as EventBus", referencing the local bus variable and the
validateActionExecutors(...) call to locate where to change.

In `@src/plugins/__tests__/CeremonyPlugin.test.ts`:
- Around line 113-160: The test "CeremonyPlugin loader: hot-reload flips
enabled→disabled (cancels timer, removes from registry)" is flaky due to very
small fixed delays (setTimeout 5ms and 10ms); either replace those sleeps with a
small polling helper (e.g., waitFor that repeatedly checks
plugin.getCeremonies() or timers.has("board.flipping") until the condition is
met) and await it after calling pluginAny._checkForChanges, or simply increase
the delays to a safer margin (e.g., 20–50ms). Update references in the test to
use the polling helper or larger timeouts around the pluginAny._checkForChanges
calls that follow writeCeremony so CI stability improves.

In `@src/plugins/CeremonyPlugin.ts`:
- Around line 538-547: The unregisterCeremony path currently removes entries
from the ceremonies and status maps but leaves per-ceremony execution data in
CeremonyStateExtension (state.history and state.lastRun), so update the
unregisterCeremony method to also clear any stored execution history and lastRun
for the given ceremony id (e.g., remove state.history[id] and state.lastRun[id]
or call the appropriate CeremonyStateExtension clear method) to avoid stale
history persisting after a hot-disable; ensure you reference and clear the same
keys used by CeremonyStateExtension.ts (state.history and state.lastRun) when
unregistering.

In `@workspace/ceremonies/health-check.yaml`:
- Around line 5-12: Summary: The ceremony is routed correctly to quinn but its
identity still reads as a service-health check while running the board_audit
skill, which is misleading. Fix: change the ceremony identity or name and/or
replace the skill to match intent so operators aren't confused — either rename
this ceremony from a "service-health" framing to something that reflects a board
audit, or swap the skill to an actual health-report skill; specifically update
the YAML entries referencing the skill value "board_audit" and any
metadata/label that implies "service-health" so the ceremony name/skill and the
identity are consistent with each other and with the target "quinn".

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 95cf24cb-d206-4df0-bbe1-850bb259bae2

📥 Commits

Reviewing files that changed from the base of the PR and between 4e910d7 and 647e1a4.

📒 Files selected for processing (45)

• .claude/commands/rollcall.md
• .env.dist
• README.md
• __tests__/goal_evaluator_plugin.test.ts
• docs/reference/ceremony-plugin.md
• docs/reference/env-vars.md
• docs/reference/http-api.md
• lib/plugins/pr-remediator.ts
• src/api/__tests__/agent-card.test.ts
• src/api/agent-card.ts
• src/config/env.ts
• src/index.ts
• src/loaders/ceremonyYamlLoader.ts
• src/planner/__tests__/end-to-end-loop.test.ts
• src/planner/__tests__/validate-action-executors.test.ts
• src/planner/types/action.ts
• src/planner/validate-action-executors.ts
• src/plugins/CeremonyPlugin.ts
• src/plugins/__tests__/CeremonyPlugin.test.ts
• src/plugins/__tests__/alert-skill-executor-plugin.test.ts
• src/plugins/__tests__/ceremony-skill-executor-plugin.test.ts
• src/plugins/__tests__/pr-remediator-skill-executor-plugin.test.ts
• src/plugins/action-dispatcher-plugin.test.ts
• src/plugins/action-dispatcher-plugin.ts
• src/plugins/agent-fleet-health-plugin.test.ts
• src/plugins/agent-fleet-health-plugin.ts
• src/plugins/alert-skill-executor-plugin.ts
• src/plugins/ceremony-skill-executor-plugin.ts
• src/plugins/goal_evaluator_plugin.ts
• src/plugins/pr-remediator-skill-executor-plugin.ts
• src/pr-remediator.test.ts
• src/schemas/yaml-schemas.ts
• src/telemetry/telemetry-service.ts
• src/world/extensions/CeremonyStateExtension.ts
• src/world/extensions/__tests__/CeremonyStateExtension.test.ts
• workspace/actions.yaml
• workspace/agents.yaml
• workspace/agents/ava.yaml
• workspace/ceremonies/agent-health.yaml
• workspace/ceremonies/board.cleanup.yaml
• workspace/ceremonies/board.health.yaml
• workspace/ceremonies/board.retro.yaml
• workspace/ceremonies/daily-standup.yaml
• workspace/ceremonies/health-check.yaml
• workspace/ceremonies/service-health.yaml