diff --git a/packages/coding-agent/src/cli/list-models.test.ts b/packages/coding-agent/src/cli/list-models.test.ts
index f35f50c04..b98493d8e 100644
--- a/packages/coding-agent/src/cli/list-models.test.ts
+++ b/packages/coding-agent/src/cli/list-models.test.ts
@@ -1,4 +1,7 @@
 import assert from "node:assert/strict";
+import { writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
 import type { Model } from "@singularity-forge/ai";
 import { afterEach, beforeEach, describe, it } from "vitest";
 import type { ModelRegistry } from "../core/model-registry.js";
@@ -17,6 +20,15 @@ const model = (provider: string, id: string): Model<any> => ({
 	maxTokens: 16000,
 });
 
+/** Minimal no-op ModelRegistry stub used by discovery-cache tests */
+const staticRegistry = (staticModels: Model<any>[] = []): ModelRegistry =>
+	({
+		discoverModels: async () => [],
+		getAvailable: () => staticModels,
+		getDiscoveredModels: () => [],
+		isDiscovered: () => false,
+	}) as unknown as ModelRegistry;
+
 let originalLog: typeof console.log;
 let output: string[];
 
@@ -86,3 +98,109 @@ describe("listModels", () => {
 		assert.match(rendered, /MiniMax-M2\.7/);
 	});
 });
+
+describe("listModels – discovery cache merge", () => {
+	/** Write a minimal discovery-cache.json to a temp file and return its path */
+	function writeTempCache(entries: Record<string, { models: object[] }>): string {
+		const path = join(tmpdir(), `discovery-cache-test-${Date.now()}-${Math.random().toString(36).slice(2)}.json`);
+		writeFileSync(
+			path,
+			JSON.stringify({ version: 1, entries }),
+			"utf-8",
+		);
+		return path;
+	}
+
+	it("includes opencode models from discovery cache when provider is absent from static catalog", async () => {
+		const cachePath = writeTempCache({
+			opencode: {
+				models: [
+					{ id: "big-pickle", name: "big-pickle", provider: "opencode", api: "openai-completions", baseUrl: "https://opencode.ai/zen", input: ["text"] },
+					{ id: "gpt-5-nano", name: "gpt-5-nano", provider: "opencode", api: "openai-completions", baseUrl: "https://opencode.ai/zen", input: ["text"] },
+				],
+			},
+		});
+
+		await listModels(staticRegistry(), {
+			discover: true,
+			_discoveryCacheFilePath: cachePath,
+		});
+
+		const rendered = output.join("\n");
+		assert.match(rendered, /opencode/);
+		assert.match(rendered, /big-pickle/);
+		assert.match(rendered, /gpt-5-nano/);
+	});
+
+	it("dedupes opencode-go: 9 static + 15 in cache → 15 unique rows (static wins for overlap)", async () => {
+		// 9 static models
+		const staticModels = Array.from({ length: 9 }, (_, i) =>
+			model("opencode-go", `static-model-${i}`),
+		);
+		// Discovery cache has 15 models: 9 overlap with static, 6 unique
+		const cacheModels = [
+			...Array.from({ length: 9 }, (_, i) => ({ id: `static-model-${i}` })),
+			...Array.from({ length: 6 }, (_, i) => ({ id: `live-only-${i}` })),
+		];
+		const cachePath = writeTempCache({
+			"opencode-go": { models: cacheModels },
+		});
+
+		await listModels(staticRegistry(staticModels), {
+			discover: true,
+			_discoveryCacheFilePath: cachePath,
+		});
+
+		const rendered = output.join("\n");
+		// Static entries should be present
+		assert.match(rendered, /static-model-0/);
+		// Live-only entries should be present
+		assert.match(rendered, /live-only-0/);
+		// Total opencode-go rows = 15 (9 static + 6 new from cache)
+		const lines = rendered.split("\n").filter((l) => l.includes("opencode-go"));
+		assert.equal(lines.length, 15, `expected 15 opencode-go rows, got ${lines.length}`);
+	});
+
+	it("gracefully handles missing discovery cache file — falls back to static only", async () => {
+		const nonExistentPath = join(tmpdir(), `no-such-file-${Date.now()}.json`);
+
+		await listModels(staticRegistry([model("anthropic", "claude-opus-4")]), {
+			discover: true,
+			_discoveryCacheFilePath: nonExistentPath,
+		});
+
+		const rendered = output.join("\n");
+		assert.match(rendered, /claude-opus-4/);
+	});
+
+	it("gracefully handles malformed discovery cache JSON — falls back to static only", async () => {
+		const path = join(tmpdir(), `bad-cache-${Date.now()}.json`);
+		writeFileSync(path, "{ this is not valid json ::::", "utf-8");
+
+		await listModels(staticRegistry([model("anthropic", "claude-sonnet-4")]), {
+			discover: true,
+			_discoveryCacheFilePath: path,
+		});
+
+		const rendered = output.join("\n");
+		assert.match(rendered, /claude-sonnet-4/);
+	});
+
+	it("synthesized cache entries are marked [discovered] in output", async () => {
+		const cachePath = writeTempCache({
+			"kimi-coding": {
+				models: [{ id: "kimi-for-coding", provider: "kimi-coding", api: "openai-completions", input: ["text"] }],
+			},
+		});
+
+		await listModels(staticRegistry(), {
+			discover: true,
+			_discoveryCacheFilePath: cachePath,
+		});
+
+		const rendered = output.join("\n");
+		const kimiLine = rendered.split("\n").find((l) => l.includes("kimi-for-coding"));
+		assert.ok(kimiLine, "kimi-for-coding should appear in output");
+		assert.match(kimiLine, /\[discovered\]/, "cache entry should be marked [discovered]");
+	});
+});
diff --git a/packages/coding-agent/src/cli/list-models.ts b/packages/coding-agent/src/cli/list-models.ts
index 8d056f7d7..2d2c39b54 100644
--- a/packages/coding-agent/src/cli/list-models.ts
+++ b/packages/coding-agent/src/cli/list-models.ts
@@ -2,11 +2,87 @@
  * List available models with optional fuzzy search and discovery support
  */
 
+import { readFileSync } from "node:fs";
+import { homedir } from "node:os";
+import { join } from "node:path";
 import type { Api, Model } from "@singularity-forge/ai";
 import { fuzzyFilter } from "@singularity-forge/tui/fuzzy";
 import { getDiscoverableProviders } from "../core/model-discovery.js";
 import type { ModelRegistry } from "../core/model-registry.js";
 
+/**
+ * Read ~/.sf/agent/discovery-cache.json and return synthesized Model objects
+ * for entries not already present in `existingKeys` (set of "provider/id" strings).
+ * Entries from the cache are marked with `_fromDiscoveryCache: true` via the
+ * badge field in the row builder below.
+ *
+ * Gracefully returns [] on missing file, malformed JSON, or unexpected shape.
+ */
+export interface SynthesizedModel extends Model<Api> {
+	/** True when this model came from the discovery cache, not the static catalog */
+	_fromDiscoveryCache: boolean;
+}
+
+function readDiscoveryCacheModels(
+	existingKeys: Set<string>,
+	cacheFilePath?: string,
+): SynthesizedModel[] {
+	try {
+		const path =
+			cacheFilePath ?? join(homedir(), ".sf", "agent", "discovery-cache.json");
+		const raw = readFileSync(path, "utf-8");
+		const data = JSON.parse(raw) as {
+			version?: number;
+			entries?: Record<
+				string,
+				{
+					models?: Array<{
+						id: string;
+						name?: string;
+						contextWindow?: number;
+						maxTokens?: number;
+						reasoning?: boolean;
+						input?: ("text" | "image")[];
+						cost?: {
+							input: number;
+							output: number;
+							cacheRead: number;
+							cacheWrite: number;
+						};
+					}>;
+				}
+			>;
+		};
+		if (data.version !== 1 || typeof data.entries !== "object") return [];
+
+		const synthesized: SynthesizedModel[] = [];
+		for (const [provider, entry] of Object.entries(data.entries)) {
+			if (!Array.isArray(entry?.models)) continue;
+			for (const m of entry.models) {
+				if (!m?.id) continue;
+				const key = `${provider}/${m.id}`;
+				if (existingKeys.has(key)) continue;
+				synthesized.push({
+					provider,
+					id: m.id,
+					name: m.name ?? m.id,
+					api: "openai-completions",
+					baseUrl: "",
+					contextWindow: m.contextWindow ?? 0,
+					maxTokens: m.maxTokens ?? 0,
+					cost: m.cost ?? { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+					reasoning: m.reasoning ?? false,
+					input: m.input ?? ["text"],
+					_fromDiscoveryCache: true,
+				});
+			}
+		}
+		return synthesized;
+	} catch {
+		return [];
+	}
+}
+
 export interface ListModelsOptions {
 	/** Include discovered models in output */
 	discover?: boolean;
@@ -19,6 +95,11 @@ export interface ListModelsOptions {
 	 * preferences.
 	 */
 	modelFilter?: (model: Model<Api>) => boolean;
+	/**
+	 * Override path to discovery-cache.json. Defaults to
+	 * ~/.sf/agent/discovery-cache.json. Intended for testing only.
+	 */
+	_discoveryCacheFilePath?: string;
 }
 
 /**
@@ -85,7 +166,7 @@ export async function listModels(
 	// Live-listed providers must not fall back to stale static catalog rows once a
 	// discovery pass was attempted. If the provider returns 401/429/empty, it
 	// contributes zero rows to this diagnostic output.
-	const models = shouldDiscover
+	const baseModels = shouldDiscover
 		? [
 				...modelRegistry
 					.getAvailable()
@@ -96,13 +177,30 @@ export async function listModels(
 			]
 		: modelRegistry.getAvailable();
 
+	// Merge entries from the discovery cache that aren't already in baseModels.
+	// This covers SF-managed providers (opencode, opencode-go, kimi-coding, …)
+	// whose SDK adapters have supportsDiscovery=false and are therefore skipped
+	// by modelRegistry.discoverModels(). The cache was populated by
+	// refreshSfManagedProviders() before this call (in src/cli.ts).
+	// Static catalog entries win over synthesized cache entries (dedup by
+	// "provider/id" — existing key = already in baseModels → skip).
+	const existingKeys = new Set(baseModels.map((m) => `${m.provider}/${m.id}`));
+	const cacheModels = readDiscoveryCacheModels(
+		existingKeys,
+		options._discoveryCacheFilePath,
+	);
+	const models: (Model<Api> | SynthesizedModel)[] = [
+		...baseModels,
+		...cacheModels,
+	];
+
 	if (models.length === 0) {
 		console.log("No models available. Set API keys in environment variables.");
 		return;
 	}
 
 	// Apply fuzzy filter if search pattern provided
-	let filteredModels: Model<Api>[] = exactDiscoveryProvider
+	let filteredModels: (Model<Api> | SynthesizedModel)[] = exactDiscoveryProvider
 		? models.filter((m) => m.provider.toLowerCase() === exactDiscoveryProvider)
 		: models;
 	if (options.searchPattern && !exactDiscoveryProvider) {
@@ -133,7 +231,9 @@ export async function listModels(
 
 	// Calculate column widths
 	const rows = filteredModels.map((m) => {
-		const isDiscovered = options.discover && modelRegistry.isDiscovered(m);
+		const isCacheEntry = "_fromDiscoveryCache" in m && m._fromDiscoveryCache;
+		const isDiscovered =
+			isCacheEntry || (options.discover && modelRegistry.isDiscovered(m));
 		return {
 			provider: m.provider,
 			model: m.id,
diff --git a/src/headless-mark-state.ts b/src/headless-mark-state.ts
new file mode 100644
index 000000000..863c51712
--- /dev/null
+++ b/src/headless-mark-state.ts
@@ -0,0 +1,214 @@
+/**
+ * Headless commands for marking SF state out-of-band:
+ *
+ *   sf headless complete-slice    <M>/<S> [--reason "..."]
+ *   sf headless skip-slice        <M>/<S> [--reason "..."]
+ *   sf headless complete-milestone <M>     [--reason "..."]
+ *
+ * Why these exist: there are several legitimate cases where an
+ * operator needs to flip slice or milestone status without going
+ * through the autonomous flow:
+ *
+ *   - Migration: pulling a planning doc into the DB-first format
+ *     leaves stale placeholder slices that wedge the active-slice
+ *     resolver. Without a CLI you'd drop to sqlite3 / python.
+ *   - Recovery: an autonomous run dies between writing the
+ *     filesystem evidence and flipping the DB row.
+ *   - Recording out-of-band work: code shipped through a parallel
+ *     channel (manual dev, paired session) and the slice needs to
+ *     reflect that without re-running execute-slice.
+ *
+ * All three commands bypass the RPC child for the same reason
+ * `headless query` does — they're deterministic DB operations that
+ * shouldn't pay the cost of spinning up an agent. They are
+ * idempotent: a slice already in the target status returns ok
+ * with a note rather than failing.
+ *
+ * Refuses cross-milestone-id ambiguity. Refuses unknown ids
+ * outright instead of silently no-op'ing (which would be a
+ * dangerous-looking 0-row UPDATE).
+ */
+
+import { existsSync } from "node:fs";
+import { join } from "node:path";
+import { createJiti } from "@mariozechner/jiti";
+import { resolveBundledSourceResource } from "./bundled-resource-path.js";
+import { getSfEnv } from "./env.js";
+
+const jiti = createJiti(import.meta.filename, {
+	interopDefault: true,
+	debug: false,
+});
+
+const agentExtensionsDir = join(getSfEnv().agentDir, "extensions", "sf");
+const useAgentDir = existsSync(join(agentExtensionsDir, "state.js"));
+
+export interface MarkStateResult {
+	exitCode: number;
+}
+
+interface MarkStateOptions {
+	command: "complete-slice" | "skip-slice" | "complete-milestone";
+	args: string[];
+	json: boolean;
+}
+
+function sfExtensionPath(moduleName: string): string {
+	if (useAgentDir) return join(agentExtensionsDir, `${moduleName}.js`);
+	const tsPath = resolveBundledSourceResource(
+		import.meta.url,
+		"extensions",
+		"sf",
+		`${moduleName}.ts`,
+	);
+	if (existsSync(tsPath)) return tsPath;
+	return resolveBundledSourceResource(
+		import.meta.url,
+		"extensions",
+		"sf",
+		`${moduleName}.js`,
+	);
+}
+
+function parseRef(ref: string): { milestoneId: string; sliceId: string | null } {
+	const trimmed = ref.trim();
+	const slash = trimmed.indexOf("/");
+	if (slash < 0) return { milestoneId: trimmed, sliceId: null };
+	return {
+		milestoneId: trimmed.slice(0, slash),
+		sliceId: trimmed.slice(slash + 1),
+	};
+}
+
+function extractReason(args: string[]): string {
+	const idx = args.indexOf("--reason");
+	if (idx >= 0 && idx + 1 < args.length) {
+		return args[idx + 1] ?? "";
+	}
+	return "";
+}
+
+function emit(
+	out: NodeJS.WritableStream,
+	json: boolean,
+	payload: Record<string, unknown>,
+	humanLine: string,
+): void {
+	if (json) {
+		out.write(`${JSON.stringify(payload)}\n`);
+	} else {
+		out.write(`${humanLine}\n`);
+	}
+}
+
+export async function handleMarkState(
+	basePath: string,
+	options: MarkStateOptions,
+): Promise<MarkStateResult> {
+	// First positional arg is the milestone/slice ref. The dispatcher in
+	// headless.ts has already filtered the flags out of args, but we still
+	// have to skip --reason and its value here.
+	const positional = options.args.filter((arg, i, all) => {
+		if (arg === "--reason") return false;
+		if (i > 0 && all[i - 1] === "--reason") return false;
+		return !arg.startsWith("--");
+	});
+	if (positional.length === 0) {
+		process.stderr.write(
+			`[headless] Error: ${options.command} requires a reference (e.g. M010/S03)\n`,
+		);
+		return { exitCode: 2 };
+	}
+	const reason = extractReason(options.args);
+	const ref = parseRef(positional[0]);
+
+	const autoStartModule = (await jiti.import(sfExtensionPath("auto-start"), {})) as {
+		openProjectDbIfPresent: (basePath: string) => Promise<void>;
+	};
+	await autoStartModule.openProjectDbIfPresent(basePath);
+
+	const slicesModule = (await jiti.import(sfExtensionPath("sf-db/sf-db-slices"), {})) as {
+		getSlice: (mid: string, sid: string) => { status: string } | null;
+		updateSliceStatus: (mid: string, sid: string, status: string, completedAt: string | null) => void;
+		setSliceSummaryMd?: (mid: string, sid: string, summaryMd: string, uatMd: string) => void;
+	};
+	const milestonesModule = (await jiti.import(sfExtensionPath("sf-db/sf-db-milestones"), {})) as {
+		getMilestone: (id: string) => { status: string } | null;
+		updateMilestoneStatus: (id: string, status: string, completedAt: string | null) => void;
+	};
+
+	const now = new Date().toISOString();
+
+	if (options.command === "complete-milestone") {
+		if (ref.sliceId !== null) {
+			process.stderr.write(
+				`[headless] Error: complete-milestone takes a milestone id, not <M>/<S>\n`,
+			);
+			return { exitCode: 2 };
+		}
+		const m = milestonesModule.getMilestone(ref.milestoneId);
+		if (!m) {
+			process.stderr.write(
+				`[headless] Error: milestone ${ref.milestoneId} not found\n`,
+			);
+			return { exitCode: 1 };
+		}
+		if (m.status === "complete") {
+			emit(process.stdout, options.json, {
+				ok: true,
+				idempotent: true,
+				milestone_id: ref.milestoneId,
+				status: m.status,
+			}, `${ref.milestoneId} already complete`);
+			return { exitCode: 0 };
+		}
+		milestonesModule.updateMilestoneStatus(ref.milestoneId, "complete", now);
+		emit(process.stdout, options.json, {
+			ok: true,
+			milestone_id: ref.milestoneId,
+			previous_status: m.status,
+			status: "complete",
+			completed_at: now,
+			reason,
+		}, `${ref.milestoneId}: ${m.status} → complete`);
+		return { exitCode: 0 };
+	}
+
+	// complete-slice / skip-slice
+	if (ref.sliceId === null) {
+		process.stderr.write(
+			`[headless] Error: ${options.command} requires <milestone>/<slice> (e.g. M010/S03)\n`,
+		);
+		return { exitCode: 2 };
+	}
+	const slice = slicesModule.getSlice(ref.milestoneId, ref.sliceId);
+	if (!slice) {
+		process.stderr.write(
+			`[headless] Error: slice ${ref.milestoneId}/${ref.sliceId} not found\n`,
+		);
+		return { exitCode: 1 };
+	}
+	const targetStatus =
+		options.command === "complete-slice" ? "complete" : "skipped";
+	if (slice.status === targetStatus) {
+		emit(process.stdout, options.json, {
+			ok: true,
+			idempotent: true,
+			milestone_id: ref.milestoneId,
+			slice_id: ref.sliceId,
+			status: slice.status,
+		}, `${ref.milestoneId}/${ref.sliceId} already ${targetStatus}`);
+		return { exitCode: 0 };
+	}
+	slicesModule.updateSliceStatus(ref.milestoneId, ref.sliceId, targetStatus, now);
+	emit(process.stdout, options.json, {
+		ok: true,
+		milestone_id: ref.milestoneId,
+		slice_id: ref.sliceId,
+		previous_status: slice.status,
+		status: targetStatus,
+		completed_at: now,
+		reason,
+	}, `${ref.milestoneId}/${ref.sliceId}: ${slice.status} → ${targetStatus}`);
+	return { exitCode: 0 };
+}
diff --git a/src/headless.ts b/src/headless.ts
index a13b7069d..73fa6705f 100644
--- a/src/headless.ts
+++ b/src/headless.ts
@@ -816,6 +816,23 @@ async function runHeadlessOnce(
 		return { exitCode: result.exitCode, interrupted: false, timedOut: false };
 	}
 
+	// Out-of-band slice/milestone status updates. Bypass the RPC child
+	// because they're deterministic single-row DB UPDATEs — no agent
+	// needed. See headless-mark-state.ts for the why.
+	if (
+		options.command === "complete-slice" ||
+		options.command === "skip-slice" ||
+		options.command === "complete-milestone"
+	) {
+		const { handleMarkState } = await import("./headless-mark-state.js");
+		const result = await handleMarkState(process.cwd(), {
+			command: options.command,
+			args: options.commandArgs,
+			json: options.json,
+		});
+		return { exitCode: result.exitCode, interrupted: false, timedOut: false };
+	}
+
 	// UOK gate health: `sf headless status uok [--json]`
 	// Bypasses the RPC path for instant, TTY-independent gate health output.
 	if (options.command === "status" && options.commandArgs[0] === "uok") {
diff --git a/src/help-text.ts b/src/help-text.ts
index a24b32391..35e5fea44 100644
--- a/src/help-text.ts
+++ b/src/help-text.ts
@@ -226,6 +226,9 @@ const SUBCOMMAND_HELP: Record<string, string> = {
 		"  usage                Live LLM-provider usage snapshot (today: gemini-cli tier + per-model quota)",
 		"  reflect              Assemble reflection corpus + render prompt for cross-corpus pattern analysis (--json for raw, --run to dispatch to gemini-cli, --model <id> to override)",
 		"  triage               Render canonical self-feedback triage prompt (--list/--json inspect, --run writes decisions, --apply runs triage-decider -> review-code)",
+		"  complete-slice       Mark a slice complete out-of-band: complete-slice <M>/<S> [--reason <txt>]",
+		"  skip-slice           Mark a slice skipped out-of-band (placeholder/migration cleanup)",
+		"  complete-milestone   Mark a milestone complete out-of-band: complete-milestone <M> [--reason <txt>]",
 		"",
 		"new-milestone flags:",
 		"  --context <path>     Path to spec/PRD file (use '-' for stdin)",
@@ -262,6 +265,9 @@ const SUBCOMMAND_HELP: Record<string, string> = {
 		"  sf headless triage --apply                     Apply via triage-decider, then gate with review-code",
 		"  sf headless reflect                            Render reflection prompt for piping",
 		"  sf headless reflect --run                      Dispatch reflection + write report",
+		"  sf headless complete-slice M010/S03            Flip M010/S03 to status=complete (idempotent)",
+		"  sf headless skip-slice M003/S01 --reason \"migration placeholder\"   Mark placeholder slice skipped",
+		"  sf headless complete-milestone M010            Flip milestone to status=complete",
 		"",
 		"Exit codes: 0 = success, 1 = error/timeout, 10 = blocked, 11 = cancelled",
 	].join("\n"),
diff --git a/src/resources/extensions/sf/bootstrap/system-context.js b/src/resources/extensions/sf/bootstrap/system-context.js
index 89c197c88..890536ab4 100644
--- a/src/resources/extensions/sf/bootstrap/system-context.js
+++ b/src/resources/extensions/sf/bootstrap/system-context.js
@@ -1,13 +1,18 @@
-import { existsSync, readdirSync, readFileSync, statSync, unlinkSync } from "node:fs";
-import { sfHome } from '../sf-home.js';
+import {
+	existsSync,
+	readdirSync,
+	readFileSync,
+	statSync,
+	unlinkSync,
+} from "node:fs";
 import { join } from "node:path";
+import { toPosixPath } from "../../shared/mod.js";
+import { isCanAskUser } from "../auto.js";
+import { getActiveAutoWorktreeContext } from "../auto-worktree.js";
 import {
 	markCmuxPromptShown,
 	shouldPromptToEnableCmux,
 } from "../cmux/index.js";
-import { toPosixPath } from "../../shared/mod.js";
-import { isCanAskUser } from "../auto.js";
-import { getActiveAutoWorktreeContext } from "../auto-worktree.js";
 import { buildCodeIntelligenceContextBlock } from "../code-intelligence.js";
 import {
 	ensureCodebaseMapFresh,
@@ -50,6 +55,7 @@ import {
 	isDbAvailable,
 	listSelfFeedbackEntries,
 } from "../sf-db.js";
+import { sfHome } from "../sf-home.js";
 import {
 	detectNewSkills,
 	formatSkillsXml,
@@ -391,13 +397,23 @@ const TACIT_SECTION_MAX_BYTES = 4096;
 // (#sf-moobj36p-ko6snt)
 const SELF_FEEDBACK_MAX_ENTRIES = 20;
 const SELF_FEEDBACK_MAX_CHARS = 4000;
-function loadSelfFeedbackBlock(_cwd) {
+/**
+ * Render unresolved self-feedback for startup context injection.
+ *
+ * Purpose: keep SF agents aware of current internal anomalies with stable entry
+ * IDs so they can inspect, triage, or resolve the exact feedback item instead
+ * of acting on an unaddressable summary.
+ *
+ * Consumer: buildBeforeAgentStartResult and system-context unit tests.
+ */
+export function loadSelfFeedbackBlock(_cwd) {
 	let entries = [];
 	if (isDbAvailable()) {
 		const rows = listSelfFeedbackEntries();
 		entries = rows
 			.filter((r) => !r["resolved_at"])
 			.map((r) => ({
+				id: r["id"] ?? "",
 				timestamp: r["ts"] ?? "",
 				kind: r["kind"] ?? "",
 				severity: r["severity"] ?? "low",
@@ -426,7 +442,10 @@ function loadSelfFeedbackBlock(_cwd) {
 	}
 	// Render compact summaries — full evidence is not injected here
 	const rows = kept
-		.map((e) => `- **${e.severity}** \`${e.kind}\` — ${e.summary}`)
+		.map(
+			(e) =>
+				`- **${e.severity}** ${e.id ? `\`${e.id}\` ` : ""}\`${e.kind}\` — ${e.summary}`,
+		)
 		.join("\n");
 	let block = `## Self-Feedback Entries (ordered by severity, ${kept.length}/${entries.length} shown)\n\n${rows}`;
 	// If still over char budget, drop from tail (lowest priority first)
@@ -436,7 +455,10 @@ function loadSelfFeedbackBlock(_cwd) {
 			block =
 				`## Self-Feedback Entries (ordered by severity, truncated)\n\n` +
 				kept
-					.map((e) => `- **${e.severity}** \`${e.kind}\` — ${e.summary}`)
+					.map(
+						(e) =>
+							`- **${e.severity}** ${e.id ? `\`${e.id}\` ` : ""}\`${e.kind}\` — ${e.summary}`,
+					)
 					.join("\n");
 		}
 	}
diff --git a/src/resources/extensions/sf/memory-extractor.js b/src/resources/extensions/sf/memory-extractor.js
index b44d589db..640f30f57 100644
--- a/src/resources/extensions/sf/memory-extractor.js
+++ b/src/resources/extensions/sf/memory-extractor.js
@@ -6,6 +6,7 @@
 
 import { readFileSync, statSync } from "node:fs";
 import { delay } from "./atomic-write.js";
+import { debugLog } from "./debug-logger.js";
 import {
 	applyConsolidationActions,
 	decayStaleMemories,
@@ -252,30 +253,50 @@ export async function extractMemoriesFromUnit(
 	llmCallFn,
 ) {
 	// Mutex guard
-	if (_extracting) return;
+	if (_extracting) {
+		debugLog("memory-extract", { phase: "skip", reason: "mutex-busy", unitKey: `${unitType}/${unitId}` });
+		return;
+	}
 	// Rate limit
 	const now = Date.now();
-	if (now - _lastExtractionTime < MIN_EXTRACTION_INTERVAL_MS) return;
+	if (now - _lastExtractionTime < MIN_EXTRACTION_INTERVAL_MS) {
+		debugLog("memory-extract", { phase: "skip", reason: "rate-limited", unitKey: `${unitType}/${unitId}`, lastExtraction: _lastExtractionTime });
+		return;
+	}
 	// Skip certain unit types
-	if (SKIP_TYPES.has(unitType)) return;
+	if (SKIP_TYPES.has(unitType)) {
+		debugLog("memory-extract", { phase: "skip", reason: "skip-type", unitType });
+		return;
+	}
 	const unitKey = `${unitType}/${unitId}`;
 	// Already processed
-	if (isUnitProcessed(unitKey)) return;
+	if (isUnitProcessed(unitKey)) {
+		debugLog("memory-extract", { phase: "skip", reason: "already-processed", unitKey });
+		return;
+	}
 	// Check file size
 	try {
 		const stat = statSync(activityFile);
-		if (stat.size < MIN_ACTIVITY_SIZE) return;
+		if (stat.size < MIN_ACTIVITY_SIZE) {
+			debugLog("memory-extract", { phase: "skip", reason: "file-too-small", unitKey, size: stat.size, min: MIN_ACTIVITY_SIZE });
+			return;
+		}
 	} catch {
+		debugLog("memory-extract", { phase: "skip", reason: "stat-failed", unitKey, file: activityFile });
 		return;
 	}
 	_extracting = true;
 	_lastExtractionTime = now;
+	debugLog("memory-extract", { phase: "start", unitKey, file: activityFile });
 	let userPrompt;
 	try {
 		// Read and parse activity file
 		const raw = readFileSync(activityFile, "utf-8");
 		const transcript = extractTranscriptFromActivity(raw);
-		if (!transcript.trim()) return;
+		if (!transcript.trim()) {
+			debugLog("memory-extract", { phase: "skip", reason: "empty-transcript", unitKey });
+			return;
+		}
 		// Redact secrets
 		const safeTranscript = redactSecrets(transcript);
 		// Get current memories for context
@@ -292,18 +313,23 @@ export async function extractMemoriesFromUnit(
 			safeTranscript,
 		);
 		// Call LLM
+		debugLog("memory-extract", { phase: "llm-call", unitKey, transcriptChars: safeTranscript.length });
 		const response = await llmCallFn(EXTRACTION_SYSTEM, userPrompt);
 		// Parse response
 		const actions = parseMemoryResponse(response);
+		debugLog("memory-extract", { phase: "parsed", unitKey, actions: actions.length });
 		// Apply actions (consolidation-path only — add/prune discipline enforced)
 		if (actions.length > 0) {
 			applyConsolidationActions(actions, unitType, unitId);
+			debugLog("memory-extract", { phase: "applied", unitKey, actions: actions.length });
 		}
 		// Decay stale memories periodically
 		decayStaleMemories(20);
 		// Mark unit as processed
 		markUnitProcessed(unitKey, activityFile);
-	} catch {
+		debugLog("memory-extract", { phase: "done", unitKey });
+	} catch (err) {
+		debugLog("memory-extract", { phase: "error", unitKey, error: err?.message || String(err) });
 		// Retry once after a brief delay
 		if (userPrompt) {
 			try {
@@ -312,7 +338,9 @@ export async function extractMemoriesFromUnit(
 				const actions2 = parseMemoryResponse(response2);
 				if (actions2.length > 0) applyConsolidationActions(actions2, unitType, unitId);
 				markUnitProcessed(unitKey, activityFile);
-			} catch {
+				debugLog("memory-extract", { phase: "retry-success", unitKey });
+			} catch (err2) {
+				debugLog("memory-extract", { phase: "retry-failed", unitKey, error: err2?.message || String(err2) });
 				// Non-fatal — memory extraction failure should never affect autonomous mode
 			}
 		}
diff --git a/src/resources/extensions/sf/tests/auto-halt-self-feedback.test.mjs b/src/resources/extensions/sf/tests/auto-halt-self-feedback.test.mjs
new file mode 100644
index 000000000..aa7e8e837
--- /dev/null
+++ b/src/resources/extensions/sf/tests/auto-halt-self-feedback.test.mjs
@@ -0,0 +1,117 @@
+import assert from "node:assert/strict";
+import {
+	existsSync,
+	mkdirSync,
+	readFileSync,
+	rmSync,
+	writeFileSync,
+} from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { afterEach, beforeEach, test, vi } from "vitest";
+import { HaltWatchdog } from "../auto/loop.js";
+import { installNotifyInterceptor } from "../bootstrap/notify-interceptor.js";
+import {
+	_resetNotificationStore,
+	initNotificationStore,
+} from "../notification-store.js";
+import {
+	closeDatabase,
+	listSelfFeedbackEntries,
+	openDatabase,
+} from "../sf-db.js";
+
+let testDir;
+
+beforeEach(() => {
+	_resetNotificationStore();
+	testDir = join(tmpdir(), `sf-halt-self-feedback-test-${Date.now()}`);
+	mkdirSync(testDir, { recursive: true });
+	writeFileSync(
+		join(testDir, "package.json"),
+		JSON.stringify({ name: "singularity-forge" }),
+	);
+	mkdirSync(join(testDir, ".sf"), { recursive: true });
+	openDatabase(join(testDir, ".sf", "sf.db"));
+	initNotificationStore(testDir);
+});
+
+afterEach(() => {
+	rmSync(testDir, { recursive: true, force: true });
+	_resetNotificationStore();
+	closeDatabase();
+	vi.useRealTimers();
+});
+
+test("HaltWatchdog.check()_when_idle_exceeds_threshold_creates_self_feedback_db_entry_and_markdown", () => {
+	vi.useFakeTimers();
+	const thresholdMs = 10;
+	const watchdog = new HaltWatchdog(testDir, thresholdMs);
+	const originalNotify = vi.fn();
+	const mockCtx = { ui: { notify: originalNotify }, basePath: testDir };
+	installNotifyInterceptor(mockCtx);
+
+	watchdog.heartbeat();
+	vi.advanceTimersByTime(thresholdMs + 5);
+	const result = watchdog.check(mockCtx, 42);
+
+	assert.equal(result.stuck, true);
+	assert.equal(originalNotify.mock.calls.length, 1);
+
+	const entries = listSelfFeedbackEntries();
+	assert.equal(entries.length, 1);
+	assert.equal(entries[0].kind, "runaway-loop:idle-halt");
+	assert.equal(entries[0].severity, "high");
+	assert.equal(entries[0].blocking, true);
+	assert.match(entries[0].summary, /iteration 42/);
+	assert.match(entries[0].summary, /no heartbeat for/);
+	assert.match(entries[0].evidence, /threshold 10ms/);
+
+	const markdownPath = join(testDir, ".sf", "SELF-FEEDBACK.md");
+	assert.equal(existsSync(markdownPath), true);
+	const markdown = readFileSync(markdownPath, "utf-8");
+	assert.match(markdown, /runaway-loop:idle-halt/);
+	assert.match(markdown, /iteration 42/);
+});
+
+test("HaltWatchdog.check()_deduplicates_within_same_idle_period", () => {
+	vi.useFakeTimers();
+	const thresholdMs = 10;
+	const watchdog = new HaltWatchdog(testDir, thresholdMs);
+	const originalNotify = vi.fn();
+	const mockCtx = { ui: { notify: originalNotify }, basePath: testDir };
+	installNotifyInterceptor(mockCtx);
+
+	watchdog.heartbeat();
+	vi.advanceTimersByTime(thresholdMs + 5);
+	watchdog.check(mockCtx, 1);
+	watchdog.check(mockCtx, 2);
+
+	assert.equal(listSelfFeedbackEntries().length, 1);
+});
+
+test("HaltWatchdog.check()_after_new_heartbeat_creates_second_entry", () => {
+	vi.useFakeTimers();
+	const thresholdMs = 10;
+	const watchdog = new HaltWatchdog(testDir, thresholdMs);
+	const originalNotify = vi.fn();
+	const mockCtx = { ui: { notify: originalNotify }, basePath: testDir };
+	installNotifyInterceptor(mockCtx);
+
+	watchdog.heartbeat();
+	vi.advanceTimersByTime(thresholdMs + 5);
+	watchdog.check(mockCtx, 1);
+
+	watchdog.heartbeat();
+	vi.advanceTimersByTime(thresholdMs + 5);
+	watchdog.check(mockCtx, 2);
+
+	const entries = listSelfFeedbackEntries();
+	assert.equal(entries.length, 2);
+	assert.equal(entries[0].kind, "runaway-loop:idle-halt");
+	assert.equal(entries[1].kind, "runaway-loop:idle-halt");
+
+	const markdownPath = join(testDir, ".sf", "SELF-FEEDBACK.md");
+	const markdown = readFileSync(markdownPath, "utf-8");
+	assert.match(markdown, /runaway-loop:idle-halt/);
+});
diff --git a/src/resources/extensions/sf/tests/system-context-self-feedback.test.mjs b/src/resources/extensions/sf/tests/system-context-self-feedback.test.mjs
new file mode 100644
index 000000000..50d1456d5
--- /dev/null
+++ b/src/resources/extensions/sf/tests/system-context-self-feedback.test.mjs
@@ -0,0 +1,30 @@
+import { describe, expect, test, vi } from "vitest";
+
+vi.mock("../sf-db.js", () => ({
+	getActiveMemories: vi.fn(() => []),
+	isDbAvailable: vi.fn(() => true),
+	listSelfFeedbackEntries: vi.fn(() => [
+		{
+			id: "sf-dogfood-fanout-001",
+			ts: "2026-05-15T14:03:00.000Z",
+			kind: "agent-friction",
+			severity: "medium",
+			summary:
+				"Dogfood run did not expose whether fan-out was skipped intentionally.",
+		},
+	]),
+}));
+
+const { loadSelfFeedbackBlock } = await import(
+	"../bootstrap/system-context.js"
+);
+
+describe("loadSelfFeedbackBlock", () => {
+	test("includes_entry_ids_so_agents_can_act_on_feedback", () => {
+		const block = loadSelfFeedbackBlock("/repo");
+
+		expect(block).toContain("sf-dogfood-fanout-001");
+		expect(block).toContain("agent-friction");
+		expect(block).toContain("Dogfood run did not expose");
+	});
+});