crewly 1.6.0 → 1.6.2

package/dist/backend/backend/src/services/agent/active-work-briefing.service.js

@@ -0,0 +1,759 @@
+/**
+ * Active Work Briefing Service
+ *
+ * Generates an authoritative-state briefing for an agent at session startup
+ * (and on-demand mid-session via the `core/get-my-active-work` skill).
+ *
+ * Today the agent recovery prompt is memory-only — `SessionMemoryService`
+ * dumps the last session summary, agent/project knowledge, daily logs, and
+ * recent learning files. That is supplementary; it does NOT capture
+ * authoritative runtime state. Symptom: an agent restarts amnesiac about
+ * Requests it owes a reply on, queued WorkItems that survived the restart,
+ * and in-flight delegations.
+ *
+ * This service mirrors `SessionMemoryService` (`generate*` + `format*`
+ * methods, singleton, soft-fail wrapping at the call site) and produces a
+ * per-agent slice of the live Request + WorkItem state. The briefing is
+ * prepended ABOVE the memory briefing in the agent's startup prompt so the
+ * first thing the agent reads is "what am I on the hook for RIGHT NOW".
+ *
+ * Sources:
+ * - {@link RequestService.listAll} — all Requests on disk
+ * - {@link TaskPoolService.getAllItems} — all WorkItems in the pool
+ *
+ * Design doc: `/tmp/agent-state-recovery-design.md` (architect output)
+ * GitHub issue: stevehuang0115/crewly#395
+ *
+ * @module services/agent/active-work-briefing.service
+ */
+import { LoggerService } from '../core/logger.service.js';
+import { ORCHESTRATOR_ROLE, ORCHESTRATOR_SESSION_NAME } from '../../constants.js';
+/**
+ * Default per-section item caps for non-orchestrator agents.
+ * Total cap = 45 items, ~6,300 chars (~1,500 tokens).
+ */
+const DEFAULT_CAPS = {
+    openRequests: 10,
+    activeWorkItems: 15,
+    pendingReviews: 5,
+    outboundDelegations: 10,
+    recentlyAutoResolved: 5,
+};
+/** Orchestrator gets system-wide visibility — caps × 3. */
+const ORCHESTRATOR_CAP_MULTIPLIER = 3;
+/**
+ * Hard ceiling on the orchestrator's formatted briefing (chars). If the
+ * formatted markdown crosses this threshold, the formatter drops description
+ * fields to keep the prompt tractable. Bounded — does not truncate items.
+ */
+const ORCHESTRATOR_OVERFLOW_CHARS = 25_000;
+/** Default recently-resolved window: 30 minutes. */
+const DEFAULT_RECENTLY_RESOLVED_WINDOW_MS = 30 * 60 * 1000;
+/** Per-item title cap to prevent unbounded growth. */
+const MAX_TITLE_CHARS = 80;
+/** Per-item description cap. */
+const MAX_DESCRIPTION_CHARS = 120;
+/**
+ * Request statuses that count as "open / on the hook" for briefing
+ * inclusion. Terminal statuses (`done`, `cancelled`) are excluded —
+ * memory recall handles "what did I close last week".
+ */
+const ACTIVE_REQUEST_STATUSES = new Set([
+    'open',
+    'ready',
+    'running',
+    'blocked',
+    'waiting_confirmation',
+]);
+/**
+ * WorkItem statuses that are still "in-flight" for the agent (target of a
+ * claim or work). Terminal statuses (`done`, `verified`, `cancelled`,
+ * `failed`, `rejected`) are excluded — they are recovered via memory.
+ */
+const ACTIVE_WORK_ITEM_STATUSES = new Set([
+    'queued',
+    'scheduled',
+    'proposed',
+    'accepted',
+    'running',
+    'blocked',
+    'escalated',
+]);
+/** Status indicating a WorkItem is awaiting verifier review. */
+const PENDING_REVIEW_STATUS = 'done_by_worker';
+// ---------------------------------------------------------------------------
+// Service
+// ---------------------------------------------------------------------------
+/**
+ * Singleton service producing per-agent active-work briefings.
+ *
+ * Mirrors the `SessionMemoryService.generate* + format*` shape so the
+ * `agent-registration.service.ts` glue can call them in parallel and
+ * concatenate the markdown.
+ *
+ * Stateless beyond singleton bookkeeping — every call hits the live
+ * `RequestService` + `TaskPoolService` instances.
+ */
+export class ActiveWorkBriefingService {
+    requestServiceFactory;
+    taskPoolFactory;
+    static instance = null;
+    logger;
+    /**
+     * Private constructor — use {@link getInstance}. The `requestServiceFactory`
+     * and `taskPoolFactory` are dependency seams for tests; production code
+     * lets {@link getInstance} provide the live singletons.
+     */
+    constructor(requestServiceFactory, taskPoolFactory) {
+        this.requestServiceFactory = requestServiceFactory;
+        this.taskPoolFactory = taskPoolFactory;
+        this.logger = LoggerService.getInstance().createComponentLogger('ActiveWorkBriefingService');
+    }
+    /**
+     * Returns the singleton instance.
+     *
+     * The `RequestService` and `TaskPoolService` modules are loaded LAZILY here
+     * (CommonJS `require` rather than top-of-file `import`) so that consumers
+     * of this module do not pay the import cost — and more importantly, do not
+     * transitively load `chokidar` via `v3-data.service.ts` — until they
+     * actually call `getInstance()`. This keeps the agent-registration test
+     * suite mockable without forcing every caller to mock the v3 pipeline.
+     */
+    static getInstance() {
+        if (!ActiveWorkBriefingService.instance) {
+            // eslint-disable-next-line @typescript-eslint/no-require-imports, @typescript-eslint/no-var-requires
+            const { RequestService } = require('../v3/request.service.js');
+            // eslint-disable-next-line @typescript-eslint/no-require-imports, @typescript-eslint/no-var-requires
+            const { TaskPoolService } = require('../task-pool/task-pool.service.js');
+            ActiveWorkBriefingService.instance = new ActiveWorkBriefingService(() => RequestService.getInstance(), () => TaskPoolService.getInstance());
+        }
+        return ActiveWorkBriefingService.instance;
+    }
+    /**
+     * Resets the singleton — for tests only.
+     */
+    static resetInstance() {
+        ActiveWorkBriefingService.instance = null;
+    }
+    /**
+     * Test-only escape hatch to swap the underlying service factories.
+     * Production code never calls this; the public surface is the singleton.
+     *
+     * @param requestServiceFactory - Returns a `listAll` provider
+     * @param taskPoolFactory - Returns a `getAllItems` provider
+     */
+    static createForTest(requestServiceFactory, taskPoolFactory) {
+        return new ActiveWorkBriefingService(requestServiceFactory, taskPoolFactory);
+    }
+    // -------------------------------------------------------------------------
+    // Internal helpers
+    // -------------------------------------------------------------------------
+    /**
+     * Compute the effective per-section caps for the requested role. Orchestrator
+     * gets caps × 3 for system-wide visibility; everyone else uses the defaults
+     * (with optional overrides from {@link GenerateOptions.caps}).
+     *
+     * @param role - Agent role (orchestrator gets the multiplier)
+     * @param overrides - Optional explicit caps that win over both defaults and
+     *   the role multiplier.
+     * @returns Effective caps for each section
+     */
+    resolveCaps(role, overrides) {
+        const isOrchestrator = role === ORCHESTRATOR_ROLE;
+        const multiplier = isOrchestrator ? ORCHESTRATOR_CAP_MULTIPLIER : 1;
+        return {
+            openRequests: overrides?.openRequests ?? DEFAULT_CAPS.openRequests * multiplier,
+            activeWorkItems: overrides?.activeWorkItems ?? DEFAULT_CAPS.activeWorkItems * multiplier,
+            pendingReviews: overrides?.pendingReviews ?? DEFAULT_CAPS.pendingReviews * multiplier,
+            outboundDelegations: overrides?.outboundDelegations ?? DEFAULT_CAPS.outboundDelegations * multiplier,
+            recentlyAutoResolved: overrides?.recentlyAutoResolved ?? DEFAULT_CAPS.recentlyAutoResolved * multiplier,
+        };
+    }
+    /**
+     * Convert a 'low' | 'normal' | 'high' priority to a sort key.
+     * Higher number = higher priority (sorted DESC).
+     *
+     * @param p - Priority value
+     * @returns Numeric sort key
+     */
+    priorityRank(p) {
+        switch (p) {
+            case 'high':
+                return 3;
+            case 'normal':
+                return 2;
+            case 'low':
+                return 1;
+            default:
+                return 0;
+        }
+    }
+    /**
+     * Compute age in hours since `createdAt`. Negative or NaN inputs are
+     * coerced to 0 — defensive against malformed timestamps.
+     *
+     * @param createdAt - ISO8601 timestamp
+     * @param nowMs - Current time in milliseconds
+     * @returns Age in hours (>= 0)
+     */
+    ageHours(createdAt, nowMs) {
+        const created = new Date(createdAt).getTime();
+        if (!Number.isFinite(created))
+            return 0;
+        const diff = nowMs - created;
+        return diff > 0 ? diff / (60 * 60 * 1000) : 0;
+    }
+    /**
+     * Read a string-typed field from a `metadata` blob, or null if missing /
+     * not a string. WorkItem.metadata is typed as `Record<string, unknown>` so
+     * every read needs this guard.
+     *
+     * @param metadata - WorkItem.metadata or undefined
+     * @param key - Field name to read
+     * @returns Trimmed string value or null
+     */
+    readMetadataString(metadata, key) {
+        if (!metadata)
+            return null;
+        const value = metadata[key];
+        if (typeof value !== 'string' || value.length === 0)
+            return null;
+        return value;
+    }
+    /**
+     * Apply the per-section cap. Returns the slice + a {@link SectionTruncation}
+     * object that the formatter renders as a "... and X more" marker.
+     *
+     * @param items - Sorted candidate rows
+     * @param cap - Per-section cap
+     * @returns Truncation metadata + the kept rows
+     */
+    applyCap(items, cap) {
+        const rows = items.slice(0, cap);
+        return {
+            rows,
+            truncation: {
+                totalCandidates: items.length,
+                cap,
+                truncated: items.length > cap,
+            },
+        };
+    }
+    // -------------------------------------------------------------------------
+    // Filter / project — Request rows
+    // -------------------------------------------------------------------------
+    /**
+     * Project Requests visible to this agent. Non-orchestrator agents only see
+     * Requests where they are the `ownerAgent`; the orchestrator sees all open
+     * Requests system-wide.
+     *
+     * @param requests - All Requests from {@link RequestService.listAll}
+     * @param sessionName - Agent session name
+     * @param role - Agent role
+     * @param nowMs - "Now" for age computation
+     * @returns Sorted candidate rows (priority DESC, then age DESC — older first)
+     */
+    projectRequests(requests, sessionName, role, nowMs) {
+        const isOrchestrator = role === ORCHESTRATOR_ROLE || sessionName === ORCHESTRATOR_SESSION_NAME;
+        const filtered = requests.filter((r) => {
+            if (!ACTIVE_REQUEST_STATUSES.has(r.status))
+                return false;
+            if (isOrchestrator)
+                return true;
+            // Non-orchestrator agents: must be on the hook (ownerAgent === sessionName).
+            return r.ownerAgent === sessionName;
+        });
+        const rows = filtered.map((r) => ({
+            id: r.id,
+            title: this.truncateTitle(r.title),
+            status: r.status,
+            priority: r.priority,
+            ageHours: this.ageHours(r.createdAt, nowMs),
+            source: r.sourceConversationItemId,
+        }));
+        rows.sort((a, b) => this.priorityRank(b.priority) - this.priorityRank(a.priority) ||
+            b.ageHours - a.ageHours);
+        return rows;
+    }
+    // -------------------------------------------------------------------------
+    // Filter / project — WorkItem rows
+    // -------------------------------------------------------------------------
+    /**
+     * Project active WorkItems for this agent: items the agent is on the hook
+     * to execute (target=me) OR items the agent has actively claimed
+     * (claimedBy=me, surfaced via `metadata.claimedBy`). Items targeted at this
+     * agent but already claimed by *another* agent are excluded — they are
+     * not actionable for the current session.
+     *
+     * Orchestrator override: include ALL active work items.
+     *
+     * @param items - All WorkItems from {@link TaskPoolService.getAllItems}
+     * @param sessionName - Agent session name
+     * @param role - Agent role
+     * @param nowMs - "Now" for age computation
+     * @returns Sorted candidate rows
+     */
+    projectActiveWorkItems(items, sessionName, role, nowMs) {
+        const isOrchestrator = role === ORCHESTRATOR_ROLE || sessionName === ORCHESTRATOR_SESSION_NAME;
+        const filtered = items.filter((wi) => {
+            if (!ACTIVE_WORK_ITEM_STATUSES.has(wi.status))
+                return false;
+            if (isOrchestrator)
+                return true;
+            const claimedBy = this.readMetadataString(wi.metadata, 'claimedBy');
+            const isTargetMe = wi.target === sessionName;
+            const isClaimedByMe = claimedBy === sessionName;
+            if (!isTargetMe && !isClaimedByMe)
+                return false;
+            // Excluded: target is me but another agent already claimed it.
+            if (isTargetMe && claimedBy && claimedBy !== sessionName)
+                return false;
+            return true;
+        });
+        const rows = filtered.map((wi) => ({
+            id: wi.id,
+            title: this.truncateTitle(wi.title),
+            status: wi.status,
+            ageHours: this.ageHours(wi.createdAt, nowMs),
+            priority: this.readPriorityFromMetadata(wi.metadata),
+            requestId: wi.requestId,
+            description: wi.description ? this.truncateDescription(wi.description) : undefined,
+        }));
+        rows.sort((a, b) => this.priorityRank(b.priority) - this.priorityRank(a.priority) ||
+            b.ageHours - a.ageHours);
+        return rows;
+    }
+    /**
+     * Pending review rows — WorkItems in `done_by_worker` where this agent
+     * is the verifier. Verifier signal = `metadata.verifier === sessionName`
+     * (TL/orchestrator pattern). For the orchestrator role we surface ALL
+     * `done_by_worker` items system-wide.
+     *
+     * @param items - All WorkItems from the pool
+     * @param sessionName - Agent session name
+     * @param role - Agent role
+     * @param nowMs - "Now" for age computation
+     * @returns Sorted candidate rows (oldest first — review backlog)
+     */
+    projectPendingReviews(items, sessionName, role, nowMs) {
+        const isOrchestrator = role === ORCHESTRATOR_ROLE || sessionName === ORCHESTRATOR_SESSION_NAME;
+        const filtered = items.filter((wi) => {
+            if (wi.status !== PENDING_REVIEW_STATUS)
+                return false;
+            if (isOrchestrator)
+                return true;
+            const verifier = this.readMetadataString(wi.metadata, 'verifier');
+            return verifier === sessionName;
+        });
+        const rows = filtered.map((wi) => ({
+            id: wi.id,
+            title: this.truncateTitle(wi.title),
+            ageHours: this.ageHours(wi.createdAt, nowMs),
+            claimedBy: this.readMetadataString(wi.metadata, 'claimedBy') ?? undefined,
+        }));
+        // Oldest first — review backlog should be drained FIFO.
+        rows.sort((a, b) => b.ageHours - a.ageHours);
+        return rows;
+    }
+    /**
+     * Outbound delegations — WorkItems this agent delegated and that haven't
+     * resolved yet. Identified by `metadata.delegatedBy === sessionName`.
+     * Forward-compatible: if upstream code does not yet populate
+     * `metadata.delegatedBy`, this section is empty.
+     *
+     * @param items - All WorkItems from the pool
+     * @param sessionName - Agent session name
+     * @param nowMs - "Now" for age computation
+     * @returns Sorted candidate rows
+     */
+    projectOutboundDelegations(items, sessionName, nowMs) {
+        const filtered = items.filter((wi) => {
+            if (!ACTIVE_WORK_ITEM_STATUSES.has(wi.status))
+                return false;
+            const delegatedBy = this.readMetadataString(wi.metadata, 'delegatedBy');
+            return delegatedBy === sessionName;
+        });
+        const rows = filtered.map((wi) => ({
+            id: wi.id,
+            title: this.truncateTitle(wi.title),
+            status: wi.status,
+            ageHours: this.ageHours(wi.createdAt, nowMs),
+            target: wi.target,
+        }));
+        rows.sort((a, b) => b.ageHours - a.ageHours);
+        return rows;
+    }
+    /**
+     * Recently auto-resolved WorkItems within the configured window
+     * (default 30 minutes). Surfaced so the agent knows what closed
+     * automatically while it was down. Identified by
+     * `metadata.slaResolvedAt` falling inside the window.
+     *
+     * @param items - All WorkItems from the pool
+     * @param sessionName - Agent session name (filters to items relevant to this agent)
+     * @param role - Agent role
+     * @param nowMs - "Now" for age computation
+     * @param windowMs - Window size in milliseconds
+     * @returns Sorted candidate rows (most recent first)
+     */
+    projectRecentlyResolved(items, sessionName, role, nowMs, windowMs) {
+        const isOrchestrator = role === ORCHESTRATOR_ROLE || sessionName === ORCHESTRATOR_SESSION_NAME;
+        const cutoff = nowMs - windowMs;
+        const filtered = items.filter((wi) => {
+            const resolvedAt = this.readMetadataString(wi.metadata, 'slaResolvedAt');
+            if (!resolvedAt)
+                return false;
+            const ts = new Date(resolvedAt).getTime();
+            if (!Number.isFinite(ts))
+                return false;
+            if (ts < cutoff)
+                return false;
+            if (isOrchestrator)
+                return true;
+            const claimedBy = this.readMetadataString(wi.metadata, 'claimedBy');
+            const delegatedBy = this.readMetadataString(wi.metadata, 'delegatedBy');
+            return (wi.target === sessionName ||
+                claimedBy === sessionName ||
+                delegatedBy === sessionName);
+        });
+        const rows = filtered.map((wi) => ({
+            id: wi.id,
+            title: this.truncateTitle(wi.title),
+            resolvedAt: this.readMetadataString(wi.metadata, 'slaResolvedAt') ?? '',
+            resolvedReason: this.readMetadataString(wi.metadata, 'slaResolvedReason') ?? undefined,
+        }));
+        rows.sort((a, b) => new Date(b.resolvedAt).getTime() - new Date(a.resolvedAt).getTime());
+        return rows;
+    }
+    /**
+     * Read a priority hint from WorkItem.metadata. Falls back to `'normal'`.
+     *
+     * @param metadata - WorkItem metadata blob
+     * @returns Briefing priority
+     */
+    readPriorityFromMetadata(metadata) {
+        const raw = this.readMetadataString(metadata, 'priority');
+        if (raw === 'high' || raw === 'normal' || raw === 'low')
+            return raw;
+        return 'normal';
+    }
+    /**
+     * Truncate a row title to {@link MAX_TITLE_CHARS}.
+     *
+     * @param title - Raw title
+     * @returns Truncated title (with ellipsis suffix if cut)
+     */
+    truncateTitle(title) {
+        if (title.length <= MAX_TITLE_CHARS)
+            return title;
+        return title.slice(0, MAX_TITLE_CHARS - 1) + '…';
+    }
+    /**
+     * Truncate a row description to {@link MAX_DESCRIPTION_CHARS}. Replaces
+     * newlines with spaces so a description never wraps the briefing layout.
+     *
+     * @param desc - Raw description
+     * @returns Truncated description
+     */
+    truncateDescription(desc) {
+        const oneLine = desc.replace(/\s+/g, ' ').trim();
+        if (oneLine.length <= MAX_DESCRIPTION_CHARS)
+            return oneLine;
+        return oneLine.slice(0, MAX_DESCRIPTION_CHARS - 1) + '…';
+    }
+    // -------------------------------------------------------------------------
+    // Public API — generate
+    // -------------------------------------------------------------------------
+    /**
+     * Generates an active-work briefing for the given agent.
+     *
+     * Pulls from {@link RequestService} + {@link TaskPoolService}, filters
+     * to the agent's slice (orchestrator gets system-wide), applies per-section
+     * caps, and returns the structured payload. The caller renders it via
+     * {@link formatBriefingAsMarkdown}.
+     *
+     * Soft-fail policy: any error from the underlying services bubbles up.
+     * The CALL SITE in `agent-registration.service.ts` wraps this in a try/catch
+     * with WARN-level logging — we deliberately do NOT swallow errors here so
+     * test scenarios can assert that the rejection occurs, and so the optional
+     * skill endpoint can return a structured 500 response.
+     *
+     * @param sessionName - Agent session name (e.g. `crewly-product-sam-dd2b46f7`)
+     * @param role - Agent role (`developer`, `orchestrator`, etc.)
+     * @param options - Optional caps + window overrides for testing
+     * @returns The structured briefing
+     *
+     * @throws Rethrows any error from {@link RequestService.listAll} or
+     *   {@link TaskPoolService.getAllItems}.
+     *
+     * @example
+     * ```typescript
+     * const briefing = await ActiveWorkBriefingService.getInstance()
+     *   .generateActiveWorkBriefing('dev-001', 'developer');
+     * const md = ActiveWorkBriefingService.getInstance()
+     *   .formatBriefingAsMarkdown(briefing);
+     * ```
+     */
+    async generateActiveWorkBriefing(sessionName, role, options = {}) {
+        const nowMs = options.nowMs ?? Date.now();
+        const windowMs = options.recentlyResolvedWindowMs ?? DEFAULT_RECENTLY_RESOLVED_WINDOW_MS;
+        const caps = this.resolveCaps(role, options.caps);
+        // Fetch in parallel — both services are independent.
+        const [requests, workItems] = await Promise.all([
+            this.requestServiceFactory().listAll(),
+            this.taskPoolFactory().getAllItems(),
+        ]);
+        const openRequestsAll = this.projectRequests(requests, sessionName, role, nowMs);
+        const activeWorkItemsAll = this.projectActiveWorkItems(workItems, sessionName, role, nowMs);
+        const pendingReviewsAll = this.projectPendingReviews(workItems, sessionName, role, nowMs);
+        const outboundDelegationsAll = this.projectOutboundDelegations(workItems, sessionName, nowMs);
+        const recentlyAutoResolvedAll = this.projectRecentlyResolved(workItems, sessionName, role, nowMs, windowMs);
+        const openRequests = this.applyCap(openRequestsAll, caps.openRequests);
+        const activeWorkItems = this.applyCap(activeWorkItemsAll, caps.activeWorkItems);
+        const pendingReviews = this.applyCap(pendingReviewsAll, caps.pendingReviews);
+        const outboundDelegations = this.applyCap(outboundDelegationsAll, caps.outboundDelegations);
+        const recentlyAutoResolved = this.applyCap(recentlyAutoResolvedAll, caps.recentlyAutoResolved);
+        const truncated = openRequests.truncation.truncated ||
+            activeWorkItems.truncation.truncated ||
+            pendingReviews.truncation.truncated ||
+            outboundDelegations.truncation.truncated ||
+            recentlyAutoResolved.truncation.truncated;
+        return {
+            openRequests: openRequests.rows,
+            activeWorkItems: activeWorkItems.rows,
+            pendingReviews: pendingReviews.rows,
+            outboundDelegations: outboundDelegations.rows,
+            recentlyAutoResolved: recentlyAutoResolved.rows,
+            truncation: {
+                openRequests: openRequests.truncation,
+                activeWorkItems: activeWorkItems.truncation,
+                pendingReviews: pendingReviews.truncation,
+                outboundDelegations: outboundDelegations.truncation,
+                recentlyAutoResolved: recentlyAutoResolved.truncation,
+            },
+            truncated,
+            generatedAt: new Date(nowMs).toISOString(),
+            totalCounts: {
+                requests: requests.length,
+                workItems: workItems.length,
+            },
+        };
+    }
+    // -------------------------------------------------------------------------
+    // Public API — format
+    // -------------------------------------------------------------------------
+    /**
+     * Format a briefing as markdown suitable for prompt injection.
+     *
+     * Output structure:
+     *
+     * ```
+     * ## Your Active Work
+     *
+     * ### Open Requests
+     * - **request-abc** — Slack inbound — _open · 2h_ *(memory: marked done — verify)*
+     * ...
+     * ### Active WorkItems
+     * - **wi-123** — Implement feature X — _running · 1h_
+     *   - desc here
+     * ...
+     * ```
+     *
+     * If a {@link MemoryHints} map is provided, contradicting memory notes are
+     * appended INLINE on each row, per design Q5/decision #2.
+     *
+     * Soft-formatting policy: empty briefing collapses to "(No active work —
+     * fresh start)". Sections with zero rows are omitted (no empty headings).
+     *
+     * @param briefing - Structured briefing payload
+     * @param memoryHints - Optional namespaced map of `req:<id>` / `wi:<id>` →
+     *   short note string
+     * @returns Markdown string
+     */
+    formatBriefingAsMarkdown(briefing, memoryHints) {
+        const totalRows = briefing.openRequests.length +
+            briefing.activeWorkItems.length +
+            briefing.pendingReviews.length +
+            briefing.outboundDelegations.length +
+            briefing.recentlyAutoResolved.length;
+        if (totalRows === 0) {
+            return '## Your Active Work\n\n(No active work — fresh start)';
+        }
+        const sections = ['## Your Active Work'];
+        // First pass — full description rendering. We may re-render in compact
+        // mode below if the orchestrator's briefing crosses the overflow ceiling.
+        sections.push(...this.renderSections(briefing, memoryHints, /* compact */ false));
+        let md = sections.join('\n\n').trim();
+        if (md.length > ORCHESTRATOR_OVERFLOW_CHARS) {
+            this.logger.debug('Active-work briefing exceeded overflow ceiling — re-rendering compact', {
+                chars: md.length,
+                ceiling: ORCHESTRATOR_OVERFLOW_CHARS,
+            });
+            const compact = ['## Your Active Work'];
+            compact.push(...this.renderSections(briefing, memoryHints, /* compact */ true));
+            md = compact.join('\n\n').trim();
+        }
+        return md;
+    }
+    /**
+     * Render the five briefing sub-sections to markdown lines. Pulled out of
+     * {@link formatBriefingAsMarkdown} so the overflow-recover path can reuse
+     * the same renderer with `compact=true` (drops descriptions to bound
+     * the orchestrator briefing).
+     *
+     * @param briefing - Briefing payload
+     * @param memoryHints - Optional contradiction notes
+     * @param compact - If true, drop description lines under WorkItem rows
+     * @returns Array of markdown blocks (joined by blank lines by caller)
+     */
+    renderSections(briefing, memoryHints, compact) {
+        const out = [];
+        if (briefing.openRequests.length > 0) {
+            out.push(this.renderOpenRequests(briefing.openRequests, briefing.truncation.openRequests, memoryHints));
+        }
+        if (briefing.activeWorkItems.length > 0) {
+            out.push(this.renderActiveWorkItems(briefing.activeWorkItems, briefing.truncation.activeWorkItems, memoryHints, compact));
+        }
+        if (briefing.pendingReviews.length > 0) {
+            out.push(this.renderPendingReviews(briefing.pendingReviews, briefing.truncation.pendingReviews, memoryHints));
+        }
+        if (briefing.outboundDelegations.length > 0) {
+            out.push(this.renderOutboundDelegations(briefing.outboundDelegations, briefing.truncation.outboundDelegations, memoryHints));
+        }
+        if (briefing.recentlyAutoResolved.length > 0) {
+            out.push(this.renderRecentlyResolved(briefing.recentlyAutoResolved, briefing.truncation.recentlyAutoResolved, memoryHints));
+        }
+        return out;
+    }
+    /**
+     * Render the truncation marker line for a section. No-op when the section
+     * was not truncated.
+     *
+     * @param truncation - Section truncation metadata
+     * @returns Marker line (empty string if not truncated)
+     */
+    renderTruncationMarker(truncation) {
+        if (!truncation.truncated)
+            return '';
+        const more = truncation.totalCandidates - truncation.cap;
+        return `\n_… and ${more} more (call \`get-my-active-work\` for full list)_`;
+    }
+    /**
+     * Append a memory-hint annotation to a row line, if a hint exists for the
+     * given key. Renders the annotation italicized so it visually subordinates
+     * to the actionable state row.
+     *
+     * @param line - Row line so far
+     * @param key - Hint lookup key (e.g. `req:abc-123`)
+     * @param hints - Memory hints map (optional)
+     * @returns The line with annotation appended, or unchanged
+     */
+    appendMemoryHint(line, key, hints) {
+        if (!hints)
+            return line;
+        const note = hints.get(key);
+        if (!note)
+            return line;
+        return `${line} *(memory: ${note})*`;
+    }
+    /** Format hours into a compact "Nh" / "Nm" string for row metadata. */
+    formatAge(ageHours) {
+        if (ageHours < 1)
+            return `${Math.max(1, Math.round(ageHours * 60))}m`;
+        return `${Math.round(ageHours)}h`;
+    }
+    /**
+     * Render the Open Requests section.
+     *
+     * @param rows - Capped rows
+     * @param truncation - Section truncation metadata
+     * @param hints - Memory hints
+     * @returns Markdown block
+     */
+    renderOpenRequests(rows, truncation, hints) {
+        const lines = ['### Open Requests'];
+        for (const row of rows) {
+            const base = `- **${row.id}** — ${row.title} — _${row.status} · ${row.priority} · ${this.formatAge(row.ageHours)}_`;
+            lines.push(this.appendMemoryHint(base, `req:${row.id}`, hints));
+        }
+        const marker = this.renderTruncationMarker(truncation);
+        return lines.join('\n') + marker;
+    }
+    /**
+     * Render the Active WorkItems section.
+     *
+     * @param rows - Capped rows
+     * @param truncation - Section truncation metadata
+     * @param hints - Memory hints
+     * @param compact - If true, drop description lines (orchestrator overflow)
+     * @returns Markdown block
+     */
+    renderActiveWorkItems(rows, truncation, hints, compact) {
+        const lines = ['### Active WorkItems'];
+        for (const row of rows) {
+            const reqRef = row.requestId ? ` · req=${row.requestId}` : '';
+            const base = `- **${row.id}** — ${row.title} — _${row.status} · ${row.priority} · ${this.formatAge(row.ageHours)}${reqRef}_`;
+            lines.push(this.appendMemoryHint(base, `wi:${row.id}`, hints));
+            if (!compact && row.description) {
+                lines.push(`  - ${row.description}`);
+            }
+        }
+        const marker = this.renderTruncationMarker(truncation);
+        return lines.join('\n') + marker;
+    }
+    /**
+     * Render the Pending Reviews section.
+     *
+     * @param rows - Capped rows
+     * @param truncation - Section truncation metadata
+     * @param hints - Memory hints
+     * @returns Markdown block
+     */
+    renderPendingReviews(rows, truncation, hints) {
+        const lines = ['### Pending Reviews'];
+        for (const row of rows) {
+            const claimedBy = row.claimedBy ? ` · by ${row.claimedBy}` : '';
+            const base = `- **${row.id}** — ${row.title} — _${this.formatAge(row.ageHours)}${claimedBy}_`;
+            lines.push(this.appendMemoryHint(base, `wi:${row.id}`, hints));
+        }
+        const marker = this.renderTruncationMarker(truncation);
+        return lines.join('\n') + marker;
+    }
+    /**
+     * Render the Outbound Delegations section.
+     *
+     * @param rows - Capped rows
+     * @param truncation - Section truncation metadata
+     * @param hints - Memory hints
+     * @returns Markdown block
+     */
+    renderOutboundDelegations(rows, truncation, hints) {
+        const lines = ['### Outbound Delegations'];
+        for (const row of rows) {
+            const target = row.target ? ` → ${row.target}` : '';
+            const base = `- **${row.id}** — ${row.title} — _${row.status} · ${this.formatAge(row.ageHours)}${target}_`;
+            lines.push(this.appendMemoryHint(base, `wi:${row.id}`, hints));
+        }
+        const marker = this.renderTruncationMarker(truncation);
+        return lines.join('\n') + marker;
+    }
+    /**
+     * Render the Recently Auto-Resolved section.
+     *
+     * @param rows - Capped rows
+     * @param truncation - Section truncation metadata
+     * @param hints - Memory hints
+     * @returns Markdown block
+     */
+    renderRecentlyResolved(rows, truncation, hints) {
+        const lines = ['### Recently Auto-Resolved (last 30min)'];
+        for (const row of rows) {
+            const reason = row.resolvedReason ? ` — ${row.resolvedReason}` : '';
+            const base = `- **${row.id}** — ${row.title}${reason}`;
+            lines.push(this.appendMemoryHint(base, `wi:${row.id}`, hints));
+        }
+        const marker = this.renderTruncationMarker(truncation);
+        return lines.join('\n') + marker;
+    }
+}
+//# sourceMappingURL=active-work-briefing.service.js.map