diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 5a46ae2..9be369e 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -30,6 +30,9 @@ jobs:
       - name: Test (vitest)
         run: npm test
 
+      - name: ProofLoop gate
+        run: node dist/cli.js gate
+
       - name: Smoke — CLI runs with zero runtime deps
         run: |
           node dist/cli.js help
diff --git a/README.md b/README.md
index 904a724..f3b5883 100644
--- a/README.md
+++ b/README.md
@@ -30,8 +30,10 @@ npx proofloop target --dir . --write-runner-plan
 `target` fetches the live URL or scans the codebase, recommends benchmark families with evidence,
 detects any already-configured benchmark/browser scripts, writes
 `.proofloop/target/latest-target-plan.json`, and can write a runnable
-`.proofloop/runner/target.plan.json`. It does not invent official scores; missing adapters and
-official scorer paths are recorded as blockers.
+`.proofloop/runner/target.plan.json`. It also writes a dated, LangChain-docs-style context page at
+`.proofloop/reports/latest.md` for the next human or coding agent to read before continuing the run.
+It does not invent official scores; missing adapters and official scorer paths are recorded as
+blockers.
 
 When `--write-browser-smoke` is provided with `--url` in a repo with `package.json`, Proof Loop
 writes `proofloop/browser/live-smoke.spec.ts` and a `proofloop:live-smoke` package script. That turns
@@ -45,7 +47,7 @@ npx proofloop init --agent auto --live  # config + manifest + agent docs + scrip
 npx proofloop doctor --json             # setup checks and fix commands
 npx proofloop manifest --dense          # compact repo status for agents
 npx proofloop ui contract --dense       # stable selectors/actions/assertions
-npx proofloop target --write-runner-plan # benchmark-family plan + runnable adapter discovery
+npx proofloop target --write-runner-plan # benchmark plan + context report + runner discovery
 npx proofloop prompt                    # kickoff prompt to paste into your coding agent
 npx proofloop this-repo --goal "proofloop my latest updates" --write-runner-plan
 npx proofloop runner run --plan proofloop.runner.json --budget-usd 100
@@ -78,6 +80,11 @@ a protected path: the gate definition is not the agent's to move.
 The CLI stays primary. `npx proofloop mcp` exposes the same compact read-only surfaces to MCP clients
 without loading broad repo context.
 
+ProofLoop also ships an Agent OS markdown pack under `docs/agent-os/`. It adapts the Room OS
+human-world-model idea into deterministic proof supervision: goals are contracts, the world model is
+the current target plus receipts/blockers, memory is mined from prior failures, and workers do not
+grade their own work.
+
 For a non-technical kickoff, tell Claude/Codex: "proofloop my latest repo" or
 "proofloop my latest updates." The agent-facing command is:
 
@@ -153,7 +160,7 @@ script. With neither, it reports `no_gate` with exit code 2. An unconfigured gat
 | `proofloop init --agent auto --live` | Add agent docs, manifest, package aliases, workflows, and rubrics. |
 | `proofloop doctor [--json]` | Report node/git/agent readiness, manifest/docs/scripts, Playwright/browser readiness, GitHub workflow, UI contracts, and fix commands. |
 | `proofloop manifest [--json\|--dense]` | Print project status: stack, commands, proof gates, workflows, UI contracts, blockers. |
-| `proofloop target [--url <url>] [--write-runner-plan] [--write-browser-smoke] [--json]` | Recommend benchmark families from a URL/codebase, detect or scaffold configured adapters, and write target/runner plan receipts. |
+| `proofloop target [--url <url>] [--write-runner-plan] [--write-browser-smoke] [--json]` | Recommend benchmark families from a URL/codebase, detect or scaffold configured adapters, and write target/runner plan receipts plus `.proofloop/reports/latest.md`. |
 | `proofloop docs agents --dense` | Print compact agent workflow instructions. |
 | `proofloop ui contract\|component <id>` | Discover stable `data-testid` and `data-proofloop` selectors. |
 | `proofloop template --list` / `proofloop template <id> --write` | List or write starter proof-loop templates. |
diff --git a/dist/cli.js b/dist/cli.js
index dd620ff..d7e692f 100644
--- a/dist/cli.js
+++ b/dist/cli.js
@@ -96,7 +96,7 @@ function usage() {
         "  charts latest              write local JSON/SVG proof charts",
         "  receipt verify --file <path>   verify app-produced proof receipts",
         "  runner run|resume|status|report   durable append-only task runner with budget and resume",
-        "  target [--url <url>] [--write-runner-plan] [--write-browser-smoke]   recommend benchmark families and write target receipt",
+        "  target [--url <url>] [--write-runner-plan] [--write-browser-smoke]   recommend benchmark families and write target/context receipts",
         "  mcp                        start the optional read-only MCP server",
         "  prompt                     print the one-prompt kickoff",
         "  this-repo [--goal <text>] [--write-runner-plan] [--run]",
@@ -211,6 +211,9 @@ function runDocsCommand(sub, options) {
         "setup=npx proofloop init --agent auto --live",
         "doctor=npx proofloop doctor --json",
         "manifest=npx proofloop manifest --dense",
+        "target=npx proofloop target --write-runner-plan",
+        "context=.proofloop/reports/latest.md",
+        "agent-os=docs/agent-os/README.md",
         "ui=npx proofloop ui contract --dense",
         "gate=npx proofloop gate",
         "resume=npx proofloop resume --dense",
diff --git a/dist/contextReport.d.ts b/dist/contextReport.d.ts
new file mode 100644
index 0000000..10fd738
--- /dev/null
+++ b/dist/contextReport.d.ts
@@ -0,0 +1,17 @@
+import type { ProofloopTargetPlan } from "./targetPlan";
+export type ProofloopContextReportResult = {
+    reportPath: string;
+    latestPath: string;
+    text: string;
+};
+export type ProofloopContextReportInputs = {
+    root: string;
+    targetPlan: ProofloopTargetPlan;
+    targetPlanPath: string;
+    runnerPlanPath?: string;
+    generatedAt?: string;
+};
+export declare function writeProofloopContextReport(inputs: ProofloopContextReportInputs): ProofloopContextReportResult;
+export declare function renderProofloopContextReport(inputs: ProofloopContextReportInputs & {
+    generatedAt: string;
+}): string;
diff --git a/dist/contextReport.js b/dist/contextReport.js
new file mode 100644
index 0000000..7fe2902
--- /dev/null
+++ b/dist/contextReport.js
@@ -0,0 +1,125 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.writeProofloopContextReport = writeProofloopContextReport;
+exports.renderProofloopContextReport = renderProofloopContextReport;
+const node_fs_1 = require("node:fs");
+const node_path_1 = require("node:path");
+const project_1 = require("./project");
+const REPORT_ROOT = ".proofloop/reports";
+function writeProofloopContextReport(inputs) {
+    const root = (0, node_path_1.resolve)(inputs.root);
+    const generatedAt = inputs.generatedAt ?? new Date().toISOString();
+    const runId = reportRunId(generatedAt);
+    const reportPath = (0, node_path_1.join)(root, REPORT_ROOT, runId, "proofloop-context.md");
+    const latestPath = (0, node_path_1.join)(root, REPORT_ROOT, "latest.md");
+    const text = renderProofloopContextReport({ ...inputs, root, generatedAt });
+    (0, node_fs_1.mkdirSync)((0, node_path_1.dirname)(reportPath), { recursive: true });
+    (0, node_fs_1.mkdirSync)((0, node_path_1.dirname)(latestPath), { recursive: true });
+    (0, node_fs_1.writeFileSync)(reportPath, text, "utf8");
+    (0, node_fs_1.writeFileSync)(latestPath, text, "utf8");
+    return { reportPath, latestPath, text };
+}
+function renderProofloopContextReport(inputs) {
+    const manifest = (0, project_1.buildProofloopProjectManifest)(inputs.root);
+    const gate = (0, project_1.buildReport)(inputs.root);
+    const target = inputs.targetPlan.target;
+    const runnerTasks = inputs.targetPlan.runnerPlan?.tasks ?? [];
+    const recommendations = inputs.targetPlan.recommendations;
+    const blockers = [...manifest.knownBlockers, ...inputs.targetPlan.blocked];
+    const sourceRows = [
+        ["target plan", inputs.targetPlanPath],
+        ...(inputs.runnerPlanPath ? [["runner plan", inputs.runnerPlanPath]] : []),
+        ["latest context report", (0, node_path_1.join)(inputs.root, REPORT_ROOT, "latest.md")],
+        ["manifest", (0, node_path_1.join)(inputs.root, ".proofloop", "manifest.json")],
+        ["gate state", (0, node_path_1.join)(inputs.root, ".proofloop", "gate-state.json")],
+    ];
+    return [
+        `# ProofLoop Context Report: ${manifest.repo.name}`,
+        "",
+        `Generated: ${inputs.generatedAt}`,
+        "",
+        "This page is deterministic. It is rendered from local files, URL fetch metadata, target-plan receipts, runner-plan receipts, and gate receipts. It is not an LLM-written claim that the app works.",
+        "",
+        "## Target",
+        "",
+        `- Kind: ${target.kind}`,
+        `- Repo root: ${target.root ?? manifest.repo.root}`,
+        `- Package: ${target.packageName ?? manifest.repo.name}`,
+        `- URL: ${target.url ?? "none"}`,
+        `- HTTP status: ${target.httpStatus ?? "not fetched"}`,
+        `- Title: ${target.title ?? "unknown"}`,
+        "",
+        "## Current Repo State",
+        "",
+        `- App: ${manifest.repo.app}`,
+        `- App reason: ${manifest.repo.appReason}`,
+        `- Stack: ${manifest.repo.stack.join(", ") || "unknown"}`,
+        `- Config: ${manifest.config.kind}${manifest.config.path ? ` (${manifest.config.path})` : ""}`,
+        `- Agent docs: ${manifest.agentInstructions.filter((entry) => entry.exists).map((entry) => entry.path).join(", ") || "missing"}`,
+        `- Workflows: ${manifest.workflows.join(", ") || "none"}`,
+        `- UI contracts: ${manifest.uiContracts.slice(0, 12).map((contract) => contract.id).join(", ") || "none"}`,
+        "",
+        "## Benchmark And Proof Targeting",
+        "",
+        "| Family | Fit | Confidence | Adapter | Scorer | Evidence |",
+        "|---|---|---:|---|---|---|",
+        ...recommendations.map((entry) => `| ${entry.id} | ${entry.fit} | ${entry.confidence.toFixed(2)} | ${entry.adapterStatus} | ${entry.officialScoreStatus} | ${escapeTable(entry.evidence.slice(0, 4).join("; "))} |`),
+        "",
+        "## Runnable Plan",
+        "",
+        runnerTasks.length
+            ? [
+                "| Task | Command | Cost estimate |",
+                "|---|---|---:|",
+                ...runnerTasks.map((task) => `| ${task.id} | ${escapeTable(task.command)} | ${money(task.estimatedCostUsd ?? 0)} |`),
+            ].join("\n")
+            : "No runner tasks were generated. Add proof checks, browser scripts, or benchmark adapters.",
+        "",
+        "## Gate Receipt",
+        "",
+        codeFence(gate.text.trim() || JSON.stringify(gate.json, null, 2), "text"),
+        "",
+        "## Not Done / Blocked",
+        "",
+        blockers.length ? blockers.map((blocker) => `- ${blocker}`).join("\n") : "- No blockers recorded by the current manifest or target plan.",
+        "",
+        "## Next Actions",
+        "",
+        inputs.targetPlan.nextActions.map((action) => `- ${action}`).join("\n"),
+        "",
+        "## Source Receipts",
+        "",
+        "| Source | Path |",
+        "|---|---|",
+        ...sourceRows.map(([label, path]) => `| ${label} | ${escapeTable(path)} |`),
+        "",
+        "## Agent Handoff",
+        "",
+        "Give this file to a coding agent together with the repo. The agent should treat the blockers above as the open work queue and should not claim done until `npx proofloop gate` or the configured runner plan passes.",
+        "",
+        codeFence([
+            "npx proofloop doctor --json",
+            "npx proofloop manifest --dense",
+            target.url ? `npx proofloop target --url ${target.url} --write-runner-plan` : "npx proofloop target --write-runner-plan",
+            inputs.runnerPlanPath ? `npx proofloop runner run --plan ${inputs.runnerPlanPath} --budget-usd 100` : "npx proofloop gate",
+            "npx proofloop report latest",
+        ].join("\n"), "bash"),
+        "",
+        "## Honesty Boundary",
+        "",
+        inputs.targetPlan.honesty,
+        "",
+    ].join("\n");
+}
+function reportRunId(iso) {
+    return iso.replace(/[:.]/g, "-").replace(/[^0-9A-Za-z_-]+/g, "-");
+}
+function escapeTable(value) {
+    return value.replace(/\|/g, "\\|").replace(/\r?\n/g, " ");
+}
+function codeFence(value, info) {
+    return ["```" + info, value, "```"].join("\n");
+}
+function money(value) {
+    return `$${value.toFixed(value < 0.01 ? 6 : 4)}`;
+}
diff --git a/dist/index.d.ts b/dist/index.d.ts
index cc1f980..22034f0 100644
--- a/dist/index.d.ts
+++ b/dist/index.d.ts
@@ -19,5 +19,6 @@ export * from "./mcp";
 export * from "./runner";
 export * from "./layeredPlan";
 export * from "./targetPlan";
+export * from "./contextReport";
 export * from "./receipts";
 export { runCli } from "./cli";
diff --git a/dist/index.js b/dist/index.js
index 783ec1e..8f0c12f 100644
--- a/dist/index.js
+++ b/dist/index.js
@@ -36,6 +36,7 @@ __exportStar(require("./mcp"), exports);
 __exportStar(require("./runner"), exports);
 __exportStar(require("./layeredPlan"), exports);
 __exportStar(require("./targetPlan"), exports);
+__exportStar(require("./contextReport"), exports);
 __exportStar(require("./receipts"), exports);
 var cli_1 = require("./cli");
 Object.defineProperty(exports, "runCli", { enumerable: true, get: function () { return cli_1.runCli; } });
diff --git a/dist/project.js b/dist/project.js
index a252474..78d00d9 100644
--- a/dist/project.js
+++ b/dist/project.js
@@ -45,6 +45,7 @@ const GENERATED_PACKAGE_SCRIPTS = {
     "proofloop:gate": "npx proofloop gate",
     "proofloop:resume": "npx proofloop resume --dense",
     "proofloop:doctor": "npx proofloop doctor --json",
+    "proofloop:target": "npx proofloop target --write-runner-plan",
     "proofloop:report": "npx proofloop report latest",
     "proofloop:charts": "npx proofloop charts latest",
 };
@@ -146,12 +147,15 @@ function buildAgentDocBlock(agent) {
         "- `npx proofloop doctor --json` reports exact missing setup checks with fix commands.",
         "- `npx proofloop manifest --dense` prints compact repo status, proof gates, workflows, and UI contracts.",
         "- `npx proofloop ui contract --dense` lists stable selectors before browser work.",
+        "- `npx proofloop target --write-runner-plan` writes target plan JSON and `.proofloop/reports/latest.md`.",
+        "- Read `docs/agent-os/README.md` for the ProofLoop Agent OS doctrine pack when present.",
         "",
         "Loop contract:",
         "- `npx proofloop this-repo --live` starts a local proof loop for this repo.",
         "- `npx proofloop gate` is the completion gate; transcript summaries are not proof.",
         "- `npx proofloop resume --dense` prints the next action after a stop or failure.",
         "- `npx proofloop report latest` summarizes the latest gate receipt.",
+        "- `.proofloop/reports/latest.md` is the dated context page to hand to another coding agent.",
         "- `npx proofloop charts latest` writes local proof charts from gate receipts.",
         "",
         "Guardrails:",
@@ -241,6 +245,7 @@ function buildProofloopProjectManifest(root) {
             init: "npx proofloop init --agent auto --live",
             doctor: "npx proofloop doctor --json",
             manifest: "npx proofloop manifest --dense",
+            target: "npx proofloop target --write-runner-plan",
             live: "npx proofloop this-repo --live",
             gate: "npx proofloop gate",
             resume: "npx proofloop resume --dense",
diff --git a/dist/targetPlan.d.ts b/dist/targetPlan.d.ts
index d2149ea..13155aa 100644
--- a/dist/targetPlan.d.ts
+++ b/dist/targetPlan.d.ts
@@ -47,6 +47,8 @@ export type ProofloopTargetResult = {
     plan: ProofloopTargetPlan;
     planPath: string;
     runnerPlanPath?: string;
+    reportPath?: string;
+    latestReportPath?: string;
 };
 export type ProofloopTargetOptions = {
     root: string;
@@ -85,5 +87,5 @@ export declare function buildProofloopTargetPlan(args: {
     generatedAt?: string;
 }): ProofloopTargetPlan;
 export declare function classifyBenchmarkFamilies(textInput: string, scripts?: Record<string, string>, hasLiveUrl?: boolean, seedEvidence?: string[]): ProofloopBenchmarkRecommendation[];
-export declare function formatProofloopTargetPlanDense(plan: ProofloopTargetPlan, planPath: string, runnerPlanPath?: string): string;
+export declare function formatProofloopTargetPlanDense(plan: ProofloopTargetPlan, planPath: string, runnerPlanPath?: string, reportPath?: string): string;
 export {};
diff --git a/dist/targetPlan.js b/dist/targetPlan.js
index 8459dbe..2e4bf93 100644
--- a/dist/targetPlan.js
+++ b/dist/targetPlan.js
@@ -7,6 +7,7 @@ exports.classifyBenchmarkFamilies = classifyBenchmarkFamilies;
 exports.formatProofloopTargetPlanDense = formatProofloopTargetPlanDense;
 const node_fs_1 = require("node:fs");
 const node_path_1 = require("node:path");
+const contextReport_1 = require("./contextReport");
 const layeredPlan_1 = require("./layeredPlan");
 const DEFAULT_TARGET_PLAN_PATH = (0, node_path_1.join)(".proofloop", "target", "latest-target-plan.json");
 const DEFAULT_TARGET_RUNNER_PLAN_PATH = (0, node_path_1.join)(".proofloop", "runner", "target.plan.json");
@@ -111,10 +112,16 @@ async function runProofloopTarget(options) {
     try {
         const result = await writeProofloopTargetPlan(options);
         if (options.json) {
-            log(JSON.stringify({ ...result.plan, planPath: result.planPath, runnerPlanPath: result.runnerPlanPath ?? null }, null, 2));
+            log(JSON.stringify({
+                ...result.plan,
+                planPath: result.planPath,
+                runnerPlanPath: result.runnerPlanPath ?? null,
+                reportPath: result.reportPath ?? null,
+                latestReportPath: result.latestReportPath ?? null,
+            }, null, 2));
         }
         else {
-            log(formatProofloopTargetPlanDense(result.plan, result.planPath, result.runnerPlanPath));
+            log(formatProofloopTargetPlanDense(result.plan, result.planPath, result.runnerPlanPath, result.reportPath));
         }
         return result;
     }
@@ -147,7 +154,15 @@ async function writeProofloopTargetPlan(options) {
         runnerPlanPath = (0, node_path_1.resolve)(root, DEFAULT_TARGET_RUNNER_PLAN_PATH);
         writeJson(runnerPlanPath, plan.runnerPlan);
     }
-    return { exitCode: 0, plan, planPath, ...(runnerPlanPath ? { runnerPlanPath } : {}) };
+    const report = (0, contextReport_1.writeProofloopContextReport)({ root, targetPlan: plan, targetPlanPath: planPath, ...(runnerPlanPath ? { runnerPlanPath } : {}) });
+    return {
+        exitCode: 0,
+        plan,
+        planPath,
+        ...(runnerPlanPath ? { runnerPlanPath } : {}),
+        reportPath: report.reportPath,
+        latestReportPath: report.latestPath,
+    };
 }
 function buildProofloopTargetPlan(args) {
     const root = (0, node_path_1.resolve)(args.root);
@@ -240,7 +255,7 @@ function classifyBenchmarkFamilies(textInput, scripts = {}, hasLiveUrl = false,
     }
     return recommendations.sort((a, b) => b.confidence - a.confidence || a.id.localeCompare(b.id));
 }
-function formatProofloopTargetPlanDense(plan, planPath, runnerPlanPath) {
+function formatProofloopTargetPlanDense(plan, planPath, runnerPlanPath, reportPath) {
     const lines = [
         "proofloop-target-plan",
         `target=${plan.target.kind}${plan.target.url ? ` url=${plan.target.url}` : ""}${plan.target.packageName ? ` package=${plan.target.packageName}` : ""}`,
@@ -248,6 +263,7 @@ function formatProofloopTargetPlanDense(plan, planPath, runnerPlanPath) {
         `officialScoreReady=${String(plan.summary.officialScoreReady)} runnerPlanReady=${String(plan.summary.runnerPlanReady)}`,
         `plan=${planPath}`,
         ...(runnerPlanPath ? [`runnerPlan=${runnerPlanPath}`] : []),
+        ...(reportPath ? [`report=${reportPath}`] : []),
     ];
     for (const rec of plan.recommendations.slice(0, 8)) {
         lines.push(`family=${rec.id} fit=${rec.fit} confidence=${rec.confidence.toFixed(2)} adapter=${rec.adapterStatus} scorer=${rec.officialScoreStatus}`);
diff --git a/docs/agent-os/README.md b/docs/agent-os/README.md
new file mode 100644
index 0000000..1f70e5a
--- /dev/null
+++ b/docs/agent-os/README.md
@@ -0,0 +1,47 @@
+# ProofLoop Agent OS Markdown Pack
+
+This pack is the human-readable operating system for ProofLoop. It gives users and coding agents a shared language for targets, claims, receipts, gates, workers, budgets, permissions, traces, and readiness.
+
+ProofLoop is not trying to imitate consciousness. It copies the useful human workflow:
+
+```mermaid
+flowchart LR
+  observe["Observe target"] --> model["Update proof world model"]
+  model --> plan["Plan checks and adapters"]
+  plan --> act["Run gates, browser smoke, or runner tasks"]
+  act --> verify["Verify receipts"]
+  verify --> memory["Write report and lessons"]
+  memory --> model
+```
+
+## Core Files
+
+- [soul.md](soul.md) - product constitution and honesty boundary.
+- [skills.md](skills.md) - verifier skills and expected inputs/outputs.
+- [world.md](world.md) - ProofLoop world model schema.
+- [loop.md](loop.md) - observe, target, plan, run, verify, report.
+- [harness.md](harness.md) - runner, gates, receipts, and retries.
+- [context.md](context.md) - what belongs in agent context.
+- [memory.md](memory.md) - what gets persisted across runs.
+- [goals.md](goals.md) - goal and claim semantics.
+- [workers.md](workers.md) - worker lifecycle for command tasks and browser smoke.
+- [delegation.md](delegation.md) - how external coding agents should be handed work.
+- [permissions.md](permissions.md) - actions that need approval.
+- [budget.md](budget.md) - cost, time, token, and task limits.
+- [visibility.md](visibility.md) - user-facing state and reports.
+- [interrupts.md](interrupts.md) - how steering changes the proof plan.
+- [collaboration.md](collaboration.md) - human, agent, and CI roles.
+- [artifact.md](artifact.md) - receipts, reports, traces, and deliverables.
+- [evals.md](evals.md) - capability and regression checks.
+- [failure-modes.md](failure-modes.md) - known ways agents fake done.
+- [trace-schema.md](trace-schema.md) - event names for trace-compatible receipts.
+- [readiness.md](readiness.md) - when ProofLoop can honestly say ready.
+- [research.md](research.md) - research grounding for world-model and context engineering.
+
+## Templates
+
+- [templates/room-template.md](templates/room-template.md)
+- [templates/skill-template.md](templates/skill-template.md)
+- [templates/worker-template.md](templates/worker-template.md)
+- [templates/eval-template.md](templates/eval-template.md)
+
diff --git a/docs/agent-os/artifact.md b/docs/agent-os/artifact.md
new file mode 100644
index 0000000..109da8b
--- /dev/null
+++ b/docs/agent-os/artifact.md
@@ -0,0 +1,17 @@
+# artifact.md
+
+Artifacts are proof-bearing outputs.
+
+Examples:
+
+- Gate state JSON.
+- Runner state JSON.
+- Runner ledger JSONL.
+- Target plan JSON.
+- Context report Markdown.
+- Browser screenshot or trace.
+- Official scorer output.
+- Cost ledger.
+
+Artifacts should be reproducible, timestamped, and tied to the command or target that produced them.
+
diff --git a/docs/agent-os/budget.md b/docs/agent-os/budget.md
new file mode 100644
index 0000000..e9963fc
--- /dev/null
+++ b/docs/agent-os/budget.md
@@ -0,0 +1,15 @@
+# budget.md
+
+Budgets make long-running proof work survivable.
+
+Budget dimensions:
+
+- Dollars.
+- Task count.
+- Runtime.
+- Model route.
+- Browser sessions.
+- External tool calls.
+
+When a budget is exhausted, the runner should stop with `blocked_budget`, write the ledger, and preserve the next queued task.
+
diff --git a/docs/agent-os/collaboration.md b/docs/agent-os/collaboration.md
new file mode 100644
index 0000000..6a6bb24
--- /dev/null
+++ b/docs/agent-os/collaboration.md
@@ -0,0 +1,12 @@
+# collaboration.md
+
+ProofLoop collaboration roles:
+
+- User: owns goal, permissions, budget, and final product judgment.
+- Coding agent: changes code and fixes failures.
+- ProofLoop: plans checks, supervises execution, protects receipts, and reports blockers.
+- CI: clean-room backstop for gate commands.
+- Hosted service: optional runner for teams that cannot run the harness locally.
+
+No role should grade its own unverified output.
+
diff --git a/docs/agent-os/context.md b/docs/agent-os/context.md
new file mode 100644
index 0000000..e6a6803
--- /dev/null
+++ b/docs/agent-os/context.md
@@ -0,0 +1,20 @@
+# context.md
+
+Context is scarce. ProofLoop gives agents compact state instead of full transcripts.
+
+## Include
+
+- Current user goal.
+- Latest target report.
+- Gate status and failing checks.
+- Target recommendations and blockers.
+- Runner state and ledger path.
+- Stable UI contracts.
+- Permission and budget limits.
+
+## Exclude
+
+- Raw long logs unless debugging a specific failure.
+- Stale reports when a newer target report exists.
+- Worker self-reports that lack receipts.
+
diff --git a/docs/agent-os/delegation.md b/docs/agent-os/delegation.md
new file mode 100644
index 0000000..beb50ed
--- /dev/null
+++ b/docs/agent-os/delegation.md
@@ -0,0 +1,14 @@
+# delegation.md
+
+Delegation means handing a clear proof context to a coding agent.
+
+The handoff packet should include:
+
+- Latest context report.
+- Exact blockers.
+- Allowed files and protected files.
+- Commands to run.
+- What receipt counts as done.
+
+Subagents may implement fixes. ProofLoop still owns the final gate.
+
diff --git a/docs/agent-os/evals.md b/docs/agent-os/evals.md
new file mode 100644
index 0000000..b3bbbaf
--- /dev/null
+++ b/docs/agent-os/evals.md
@@ -0,0 +1,17 @@
+# evals.md
+
+ProofLoop evals test verifier behavior.
+
+Core eval families:
+
+- Empty gate is not a pass.
+- Failing check blocks done.
+- Forged gate pass is blocked by protected paths.
+- Tool-use deny-list fails closed on missing logs.
+- Target planner labels missing adapters.
+- Browser smoke is separate from official benchmark scoring.
+- Runner resumes after interruption.
+- Watch/dev scripts are not emitted as proof tasks.
+
+When an agent discovers a new failure mode, add a test or template blocker.
+
diff --git a/docs/agent-os/failure-modes.md b/docs/agent-os/failure-modes.md
new file mode 100644
index 0000000..243d057
--- /dev/null
+++ b/docs/agent-os/failure-modes.md
@@ -0,0 +1,16 @@
+# failure-modes.md
+
+Known failure modes:
+
+- Claiming done from transcript text.
+- Editing `.proofloop/gate-state.json` to forge a pass.
+- Weakening `proofloop.config.json`.
+- Removing CI gate workflows.
+- Treating a browser smoke test as an official benchmark score.
+- Treating proxy proof as official score proof.
+- Certifying an empty tool-use log.
+- Hiding blocked adapters behind a green summary.
+- Emitting non-terminating watch tasks into a runner plan.
+
+The fix is not better wording. The fix is a gate, guard, receipt, test, or blocker.
+
diff --git a/docs/agent-os/goals.md b/docs/agent-os/goals.md
new file mode 100644
index 0000000..813aebd
--- /dev/null
+++ b/docs/agent-os/goals.md
@@ -0,0 +1,13 @@
+# goals.md
+
+ProofLoop goals are proof contracts.
+
+Examples:
+
+- "Run the repo's existing test suite and block done if it fails."
+- "Classify the app and produce benchmark-family recommendations."
+- "Scaffold a live URL smoke test and include it in the runner plan."
+- "Verify official benchmark score artifacts with the upstream scorer."
+
+Goal status must be derived from receipts: passed, failed, no_gate, blocked, or not_configured.
+
diff --git a/docs/agent-os/harness.md b/docs/agent-os/harness.md
new file mode 100644
index 0000000..59be6a8
--- /dev/null
+++ b/docs/agent-os/harness.md
@@ -0,0 +1,15 @@
+# harness.md
+
+The harness is the part that reality-checks agent work.
+
+## Harness Pieces
+
+- `proofloop gate`: deterministic completion gate.
+- `proofloop target`: target scan, benchmark matching, runner-plan discovery, and context report generation.
+- `proofloop runner`: durable command execution with ledger, locks, budget, resume, and secret redaction.
+- `proofloop hooks`: stop gate and protected path guards.
+- `proofloop tooluse`: expected tool-use contracts.
+- `proofloop receipt verify`: app-produced receipt verification.
+
+Harness tasks must terminate. Dev servers, watch mode, preview servers, and interactive prompts are not proof tasks.
+
diff --git a/docs/agent-os/interrupts.md b/docs/agent-os/interrupts.md
new file mode 100644
index 0000000..673c797
--- /dev/null
+++ b/docs/agent-os/interrupts.md
@@ -0,0 +1,14 @@
+# interrupts.md
+
+Interrupts are user steering events.
+
+Examples:
+
+- "Also test the live URL."
+- "Keep it under $100."
+- "Use free models first."
+- "Do not use memory mode."
+- "Publish to npm after CI."
+
+ProofLoop should convert interrupts into target-plan changes, runner budget changes, permission blockers, or new receipts. It should not overwrite prior goals unless the user explicitly replaces them.
+
diff --git a/docs/agent-os/loop.md b/docs/agent-os/loop.md
new file mode 100644
index 0000000..53605ce
--- /dev/null
+++ b/docs/agent-os/loop.md
@@ -0,0 +1,15 @@
+# loop.md
+
+ProofLoop runs a verification loop:
+
+1. Observe the repo and optional live URL.
+2. Build or refresh the target plan.
+3. Identify benchmark families, scripts, gates, UI contracts, and blockers.
+4. Generate or reuse runner tasks.
+5. Run deterministic gates or supervised runner tasks.
+6. Verify receipts.
+7. Write charts, reports, and next actions.
+8. Keep unresolved blockers visible.
+
+The loop is allowed to continue work. It is not allowed to mark its own work done without proof.
+
diff --git a/docs/agent-os/memory.md b/docs/agent-os/memory.md
new file mode 100644
index 0000000..b2ccd2b
--- /dev/null
+++ b/docs/agent-os/memory.md
@@ -0,0 +1,17 @@
+# memory.md
+
+ProofLoop memory is receipt memory.
+
+## Durable Memory
+
+- `.proofloop/gate-state.json`
+- `.proofloop/target/latest-target-plan.json`
+- `.proofloop/runner/runs/*/state.json`
+- `.proofloop/runner/runs/*/ledger.jsonl`
+- `.proofloop/reports/latest.md`
+- `.proofloop/charts/latest.json`
+
+## Consolidation Rule
+
+If the same failure repeats, turn it into a rule, test, protected path, target blocker, or runner task. Do not leave it only in chat history.
+
diff --git a/docs/agent-os/permissions.md b/docs/agent-os/permissions.md
new file mode 100644
index 0000000..b4620a5
--- /dev/null
+++ b/docs/agent-os/permissions.md
@@ -0,0 +1,21 @@
+# permissions.md
+
+ProofLoop distinguishes local proof from external side effects.
+
+## Usually Allowed
+
+- Read local repo files.
+- Write `.proofloop/` generated reports and receipts.
+- Run local deterministic commands.
+- Fetch a user-provided HTTP or HTTPS URL for target scanning.
+
+## Requires Care
+
+- Publishing packages.
+- Sending emails or form submissions.
+- Changing DNS, deployment settings, or secrets.
+- Running official scorers that call paid models.
+- Uploading private code or artifacts to a hosted service.
+
+ProofLoop should record permission blockers instead of pretending the action happened.
+
diff --git a/docs/agent-os/readiness.md b/docs/agent-os/readiness.md
new file mode 100644
index 0000000..e9ac252
--- /dev/null
+++ b/docs/agent-os/readiness.md
@@ -0,0 +1,16 @@
+# readiness.md
+
+ProofLoop readiness means the proof story is complete enough to state honestly.
+
+Checklist:
+
+- `npx proofloop doctor --json` has no unhandled setup blockers.
+- `npx proofloop target --write-runner-plan` wrote a target plan and context report.
+- All generated runner tasks terminate.
+- `npx proofloop gate` passes or the runner report shows passed.
+- Browser claims have browser receipts.
+- Official benchmark claims have official scorer outputs or an explicitly recorded equivalent judge contract.
+- Remaining blockers are listed in the latest context report.
+
+If any item is missing, say what is not done.
+
diff --git a/docs/agent-os/research.md b/docs/agent-os/research.md
new file mode 100644
index 0000000..ead5bf4
--- /dev/null
+++ b/docs/agent-os/research.md
@@ -0,0 +1,12 @@
+# research.md
+
+Research grounding for this pack:
+
+- World Models: agents benefit from compact spatial and temporal representations that support simulation before action. https://arxiv.org/abs/1803.10122
+- Generative Agents: believable behavior uses observation, memory, reflection, planning, and action. https://arxiv.org/abs/2304.03442
+- LangChain context engineering: agents need the right information at the right lifecycle step. https://docs.langchain.com/oss/python/langchain/context-engineering
+- Anthropic effective agents: reliable agents need clear success criteria, feedback loops, and human oversight. https://www.anthropic.com/engineering/building-effective-agents
+- Anthropic context engineering: long-running agents need curated context, not raw transcript stuffing. https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents
+
+ProofLoop adapts these ideas to verification: world state is proof state, memory is receipts, and feedback is deterministic gates plus external scorer output.
+
diff --git a/docs/agent-os/skills.md b/docs/agent-os/skills.md
new file mode 100644
index 0000000..2ee4a0a
--- /dev/null
+++ b/docs/agent-os/skills.md
@@ -0,0 +1,15 @@
+# skills.md
+
+ProofLoop skills are verifier skills. They are not personality traits.
+
+| Skill | Input | Output | Failure mode |
+|---|---|---|---|
+| Target scan | Repo and optional URL | Target plan JSON and context report | Overclaiming benchmark readiness |
+| Gate run | `proofloop.config.json` or npm test fallback | Gate receipt | Empty checks treated as pass |
+| Browser smoke | URL and Playwright repo | Smoke spec and runner task | Treating smoke as full user workflow |
+| Runner supervision | Runner plan | Append-only ledger and state | Non-terminating watch tasks |
+| Tool-use verification | Contract and log | Pass/fail contract receipt | Empty log certified as safe |
+| Context report | Target plan, manifest, gate state | Markdown handoff page | LLM prose inventing state |
+
+Every skill must expose inputs, outputs, receipts, and blocker language.
+
diff --git a/docs/agent-os/soul.md b/docs/agent-os/soul.md
new file mode 100644
index 0000000..60af06c
--- /dev/null
+++ b/docs/agent-os/soul.md
@@ -0,0 +1,17 @@
+# soul.md
+
+ProofLoop's soul is simple: the gate decides when work is done.
+
+## Principles
+
+- A self-report is not proof.
+- A screenshot is useful evidence, but not a gate by itself.
+- Product-path proof, proxy benchmark proof, and official scorer output are separate claims.
+- A verifier must not let the worker weaken the verifier.
+- Every scan should produce a dated, portable context report.
+- Every blocker should remain visible until a deterministic gate, scorer, or receipt closes it.
+
+## Product Boundary
+
+ProofLoop may schedule commands, generate plans, write reports, and scaffold smoke adapters. It does not magically know an app's official benchmark score unless the app provides the official scorer or a recorded equivalent judge contract.
+
diff --git a/docs/agent-os/templates/eval-template.md b/docs/agent-os/templates/eval-template.md
new file mode 100644
index 0000000..2f6fc53
--- /dev/null
+++ b/docs/agent-os/templates/eval-template.md
@@ -0,0 +1,18 @@
+# Eval Template
+
+## Eval
+
+Name:
+
+## Setup
+
+## Command
+
+```bash
+
+```
+
+## Expected Result
+
+## Regression Protected
+
diff --git a/docs/agent-os/templates/room-template.md b/docs/agent-os/templates/room-template.md
new file mode 100644
index 0000000..7c978be
--- /dev/null
+++ b/docs/agent-os/templates/room-template.md
@@ -0,0 +1,22 @@
+# Room / Target Template
+
+## Target
+
+- Repo:
+- URL:
+- Environment:
+- Owner:
+
+## Claims To Prove
+
+- Claim:
+- Required receipt:
+- Blocker:
+
+## Commands
+
+```bash
+npx proofloop target --url <url> --write-runner-plan
+npx proofloop gate
+```
+
diff --git a/docs/agent-os/templates/skill-template.md b/docs/agent-os/templates/skill-template.md
new file mode 100644
index 0000000..d25c3df
--- /dev/null
+++ b/docs/agent-os/templates/skill-template.md
@@ -0,0 +1,22 @@
+# Skill Template
+
+## Skill
+
+Name:
+
+## Inputs
+
+- 
+
+## Outputs
+
+- 
+
+## Receipts
+
+- 
+
+## Failure Modes
+
+- 
+
diff --git a/docs/agent-os/templates/worker-template.md b/docs/agent-os/templates/worker-template.md
new file mode 100644
index 0000000..6e995b3
--- /dev/null
+++ b/docs/agent-os/templates/worker-template.md
@@ -0,0 +1,20 @@
+# Worker Template
+
+## Worker
+
+Id:
+
+## Command
+
+```bash
+
+```
+
+## Timeout
+
+## Budget
+
+## Expected Receipt
+
+## Retry Policy
+
diff --git a/docs/agent-os/trace-schema.md b/docs/agent-os/trace-schema.md
new file mode 100644
index 0000000..00a2c44
--- /dev/null
+++ b/docs/agent-os/trace-schema.md
@@ -0,0 +1,22 @@
+# trace-schema.md
+
+Recommended trace events:
+
+- `target_scan_started`
+- `target_scan_completed`
+- `benchmark_family_matched`
+- `adapter_configured`
+- `adapter_missing`
+- `runner_plan_written`
+- `context_report_written`
+- `gate_started`
+- `gate_completed`
+- `task_started`
+- `task_completed`
+- `budget_blocked`
+- `receipt_verified`
+- `permission_required`
+- `claim_rejected`
+
+Each event should include timestamp, target id, command or receipt path, status, and blocker text when relevant.
+
diff --git a/docs/agent-os/visibility.md b/docs/agent-os/visibility.md
new file mode 100644
index 0000000..fdad676
--- /dev/null
+++ b/docs/agent-os/visibility.md
@@ -0,0 +1,15 @@
+# visibility.md
+
+Users need to see the state that drives decisions.
+
+Visible surfaces:
+
+- Target plan JSON for machines.
+- Context report Markdown for humans and agents.
+- Runner report for task, family, model, and cost summaries.
+- Gate receipt for pass/fail state.
+- Charts for quick inspection.
+- Blocker lists that persist until closed.
+
+The UI or CLI should never hide missing official scorer status behind a green-looking summary.
+
diff --git a/docs/agent-os/workers.md b/docs/agent-os/workers.md
new file mode 100644
index 0000000..1437dfb
--- /dev/null
+++ b/docs/agent-os/workers.md
@@ -0,0 +1,15 @@
+# workers.md
+
+ProofLoop workers are supervised tasks, not autonomous personalities.
+
+## Lifecycle
+
+- queued
+- running
+- passed
+- failed
+- blocked_budget
+- paused
+
+Each worker needs an id, command, working directory, optional environment, timeout, and cost estimate. The runner writes state before and after each task so the run can resume after interruption.
+
diff --git a/docs/agent-os/world.md b/docs/agent-os/world.md
new file mode 100644
index 0000000..87020ae
--- /dev/null
+++ b/docs/agent-os/world.md
@@ -0,0 +1,27 @@
+# world.md
+
+ProofLoop's world model describes proof reality, not application business logic.
+
+```ts
+type ProofWorld = {
+  target: Repo | LiveUrl | HybridTarget;
+  claims: Claim[];
+  gates: Gate[];
+  receipts: Receipt[];
+  tools: ToolAffordance[];
+  budgets: Budget[];
+  permissions: Permission[];
+  blockers: Blocker[];
+  risks: Risk[];
+  reports: ContextReport[];
+};
+```
+
+## Core Entities
+
+- Target: the codebase, live URL, or both.
+- Claim: a statement someone wants to make, such as "live browser verified" or "official score ready".
+- Receipt: machine-readable proof produced by a gate, runner, browser test, scorer, or tool-use contract.
+- Blocker: a known missing adapter, missing scorer, failing check, missing permission, or exhausted budget.
+- Report: dated Markdown snapshot that a human or agent can read.
+
diff --git a/package.json b/package.json
index 2d430ad..d9db0b4 100644
--- a/package.json
+++ b/package.json
@@ -28,6 +28,7 @@
   "files": [
     "dist",
     "templates",
+    "docs/agent-os",
     "README.md",
     "LICENSE"
   ],
@@ -37,7 +38,15 @@
     "prepublishOnly": "npm run build",
     "pretest": "npm run build",
     "test": "vitest run",
-    "test:watch": "vitest"
+    "test:watch": "vitest",
+    "proofloop:init": "npx proofloop init --agent auto --live",
+    "proofloop:live": "npx proofloop this-repo --live",
+    "proofloop:gate": "npx proofloop gate",
+    "proofloop:resume": "npx proofloop resume --dense",
+    "proofloop:doctor": "npx proofloop doctor --json",
+    "proofloop:target": "npx proofloop target --write-runner-plan",
+    "proofloop:report": "npx proofloop report latest",
+    "proofloop:charts": "npx proofloop charts latest"
   },
   "devDependencies": {
     "@webcontainer/api": "^1.6.4",
diff --git a/proofloop.config.json b/proofloop.config.json
new file mode 100644
index 0000000..85802c6
--- /dev/null
+++ b/proofloop.config.json
@@ -0,0 +1,22 @@
+{
+  "app": "ProofLoop CLI and static site",
+  "workflow": "Scan a repo or live URL, produce deterministic proof receipts, generate the context handoff report, and refuse fake done through the gate.",
+  "gate": {
+    "checks": [
+      {
+        "name": "build",
+        "command": "npm run build"
+      },
+      {
+        "name": "tests",
+        "command": "npm test"
+      }
+    ]
+  },
+  "immutable": [
+    "templates/github-proofloop-gate.yml"
+  ],
+  "protectedPaths": [
+    "docs/agent-os/"
+  ]
+}
diff --git a/src/cli.ts b/src/cli.ts
index 2f806e0..29c0953 100644
--- a/src/cli.ts
+++ b/src/cli.ts
@@ -112,7 +112,7 @@ function usage(): string {
     "  charts latest              write local JSON/SVG proof charts",
     "  receipt verify --file <path>   verify app-produced proof receipts",
     "  runner run|resume|status|report   durable append-only task runner with budget and resume",
-    "  target [--url <url>] [--write-runner-plan] [--write-browser-smoke]   recommend benchmark families and write target receipt",
+    "  target [--url <url>] [--write-runner-plan] [--write-browser-smoke]   recommend benchmark families and write target/context receipts",
     "  mcp                        start the optional read-only MCP server",
     "  prompt                     print the one-prompt kickoff",
     "  this-repo [--goal <text>] [--write-runner-plan] [--run]",
@@ -251,6 +251,9 @@ function runDocsCommand(sub: string | undefined, options: Record<string, string
     "setup=npx proofloop init --agent auto --live",
     "doctor=npx proofloop doctor --json",
     "manifest=npx proofloop manifest --dense",
+    "target=npx proofloop target --write-runner-plan",
+    "context=.proofloop/reports/latest.md",
+    "agent-os=docs/agent-os/README.md",
     "ui=npx proofloop ui contract --dense",
     "gate=npx proofloop gate",
     "resume=npx proofloop resume --dense",
diff --git a/src/contextReport.ts b/src/contextReport.ts
new file mode 100644
index 0000000..fa92b6d
--- /dev/null
+++ b/src/contextReport.ts
@@ -0,0 +1,144 @@
+import { mkdirSync, writeFileSync } from "node:fs";
+import { dirname, join, resolve } from "node:path";
+import { buildProofloopProjectManifest, buildReport } from "./project";
+import type { ProofloopTargetPlan } from "./targetPlan";
+
+export type ProofloopContextReportResult = {
+  reportPath: string;
+  latestPath: string;
+  text: string;
+};
+
+export type ProofloopContextReportInputs = {
+  root: string;
+  targetPlan: ProofloopTargetPlan;
+  targetPlanPath: string;
+  runnerPlanPath?: string;
+  generatedAt?: string;
+};
+
+const REPORT_ROOT = ".proofloop/reports";
+
+export function writeProofloopContextReport(inputs: ProofloopContextReportInputs): ProofloopContextReportResult {
+  const root = resolve(inputs.root);
+  const generatedAt = inputs.generatedAt ?? new Date().toISOString();
+  const runId = reportRunId(generatedAt);
+  const reportPath = join(root, REPORT_ROOT, runId, "proofloop-context.md");
+  const latestPath = join(root, REPORT_ROOT, "latest.md");
+  const text = renderProofloopContextReport({ ...inputs, root, generatedAt });
+  mkdirSync(dirname(reportPath), { recursive: true });
+  mkdirSync(dirname(latestPath), { recursive: true });
+  writeFileSync(reportPath, text, "utf8");
+  writeFileSync(latestPath, text, "utf8");
+  return { reportPath, latestPath, text };
+}
+
+export function renderProofloopContextReport(inputs: ProofloopContextReportInputs & { generatedAt: string }): string {
+  const manifest = buildProofloopProjectManifest(inputs.root);
+  const gate = buildReport(inputs.root);
+  const target = inputs.targetPlan.target;
+  const runnerTasks = inputs.targetPlan.runnerPlan?.tasks ?? [];
+  const recommendations = inputs.targetPlan.recommendations;
+  const blockers = [...manifest.knownBlockers, ...inputs.targetPlan.blocked];
+  const sourceRows = [
+    ["target plan", inputs.targetPlanPath],
+    ...(inputs.runnerPlanPath ? [["runner plan", inputs.runnerPlanPath] as [string, string]] : []),
+    ["latest context report", join(inputs.root, REPORT_ROOT, "latest.md")],
+    ["manifest", join(inputs.root, ".proofloop", "manifest.json")],
+    ["gate state", join(inputs.root, ".proofloop", "gate-state.json")],
+  ];
+
+  return [
+    `# ProofLoop Context Report: ${manifest.repo.name}`,
+    "",
+    `Generated: ${inputs.generatedAt}`,
+    "",
+    "This page is deterministic. It is rendered from local files, URL fetch metadata, target-plan receipts, runner-plan receipts, and gate receipts. It is not an LLM-written claim that the app works.",
+    "",
+    "## Target",
+    "",
+    `- Kind: ${target.kind}`,
+    `- Repo root: ${target.root ?? manifest.repo.root}`,
+    `- Package: ${target.packageName ?? manifest.repo.name}`,
+    `- URL: ${target.url ?? "none"}`,
+    `- HTTP status: ${target.httpStatus ?? "not fetched"}`,
+    `- Title: ${target.title ?? "unknown"}`,
+    "",
+    "## Current Repo State",
+    "",
+    `- App: ${manifest.repo.app}`,
+    `- App reason: ${manifest.repo.appReason}`,
+    `- Stack: ${manifest.repo.stack.join(", ") || "unknown"}`,
+    `- Config: ${manifest.config.kind}${manifest.config.path ? ` (${manifest.config.path})` : ""}`,
+    `- Agent docs: ${manifest.agentInstructions.filter((entry) => entry.exists).map((entry) => entry.path).join(", ") || "missing"}`,
+    `- Workflows: ${manifest.workflows.join(", ") || "none"}`,
+    `- UI contracts: ${manifest.uiContracts.slice(0, 12).map((contract) => contract.id).join(", ") || "none"}`,
+    "",
+    "## Benchmark And Proof Targeting",
+    "",
+    "| Family | Fit | Confidence | Adapter | Scorer | Evidence |",
+    "|---|---|---:|---|---|---|",
+    ...recommendations.map((entry) => `| ${entry.id} | ${entry.fit} | ${entry.confidence.toFixed(2)} | ${entry.adapterStatus} | ${entry.officialScoreStatus} | ${escapeTable(entry.evidence.slice(0, 4).join("; "))} |`),
+    "",
+    "## Runnable Plan",
+    "",
+    runnerTasks.length
+      ? [
+          "| Task | Command | Cost estimate |",
+          "|---|---|---:|",
+          ...runnerTasks.map((task) => `| ${task.id} | ${escapeTable(task.command)} | ${money(task.estimatedCostUsd ?? 0)} |`),
+        ].join("\n")
+      : "No runner tasks were generated. Add proof checks, browser scripts, or benchmark adapters.",
+    "",
+    "## Gate Receipt",
+    "",
+    codeFence(gate.text.trim() || JSON.stringify(gate.json, null, 2), "text"),
+    "",
+    "## Not Done / Blocked",
+    "",
+    blockers.length ? blockers.map((blocker) => `- ${blocker}`).join("\n") : "- No blockers recorded by the current manifest or target plan.",
+    "",
+    "## Next Actions",
+    "",
+    inputs.targetPlan.nextActions.map((action) => `- ${action}`).join("\n"),
+    "",
+    "## Source Receipts",
+    "",
+    "| Source | Path |",
+    "|---|---|",
+    ...sourceRows.map(([label, path]) => `| ${label} | ${escapeTable(path)} |`),
+    "",
+    "## Agent Handoff",
+    "",
+    "Give this file to a coding agent together with the repo. The agent should treat the blockers above as the open work queue and should not claim done until `npx proofloop gate` or the configured runner plan passes.",
+    "",
+    codeFence([
+      "npx proofloop doctor --json",
+      "npx proofloop manifest --dense",
+      target.url ? `npx proofloop target --url ${target.url} --write-runner-plan` : "npx proofloop target --write-runner-plan",
+      inputs.runnerPlanPath ? `npx proofloop runner run --plan ${inputs.runnerPlanPath} --budget-usd 100` : "npx proofloop gate",
+      "npx proofloop report latest",
+    ].join("\n"), "bash"),
+    "",
+    "## Honesty Boundary",
+    "",
+    inputs.targetPlan.honesty,
+    "",
+  ].join("\n");
+}
+
+function reportRunId(iso: string): string {
+  return iso.replace(/[:.]/g, "-").replace(/[^0-9A-Za-z_-]+/g, "-");
+}
+
+function escapeTable(value: string): string {
+  return value.replace(/\|/g, "\\|").replace(/\r?\n/g, " ");
+}
+
+function codeFence(value: string, info: string): string {
+  return ["```" + info, value, "```"].join("\n");
+}
+
+function money(value: number): string {
+  return `$${value.toFixed(value < 0.01 ? 6 : 4)}`;
+}
diff --git a/src/index.ts b/src/index.ts
index cc1f980..22034f0 100644
--- a/src/index.ts
+++ b/src/index.ts
@@ -19,5 +19,6 @@ export * from "./mcp";
 export * from "./runner";
 export * from "./layeredPlan";
 export * from "./targetPlan";
+export * from "./contextReport";
 export * from "./receipts";
 export { runCli } from "./cli";
diff --git a/src/project.ts b/src/project.ts
index b592c28..bfe5c27 100644
--- a/src/project.ts
+++ b/src/project.ts
@@ -41,6 +41,7 @@ const GENERATED_PACKAGE_SCRIPTS: Record<string, string> = {
   "proofloop:gate": "npx proofloop gate",
   "proofloop:resume": "npx proofloop resume --dense",
   "proofloop:doctor": "npx proofloop doctor --json",
+  "proofloop:target": "npx proofloop target --write-runner-plan",
   "proofloop:report": "npx proofloop report latest",
   "proofloop:charts": "npx proofloop charts latest",
 };
@@ -182,12 +183,15 @@ function buildAgentDocBlock(agent: ProofloopAgentKind): string {
     "- `npx proofloop doctor --json` reports exact missing setup checks with fix commands.",
     "- `npx proofloop manifest --dense` prints compact repo status, proof gates, workflows, and UI contracts.",
     "- `npx proofloop ui contract --dense` lists stable selectors before browser work.",
+    "- `npx proofloop target --write-runner-plan` writes target plan JSON and `.proofloop/reports/latest.md`.",
+    "- Read `docs/agent-os/README.md` for the ProofLoop Agent OS doctrine pack when present.",
     "",
     "Loop contract:",
     "- `npx proofloop this-repo --live` starts a local proof loop for this repo.",
     "- `npx proofloop gate` is the completion gate; transcript summaries are not proof.",
     "- `npx proofloop resume --dense` prints the next action after a stop or failure.",
     "- `npx proofloop report latest` summarizes the latest gate receipt.",
+    "- `.proofloop/reports/latest.md` is the dated context page to hand to another coding agent.",
     "- `npx proofloop charts latest` writes local proof charts from gate receipts.",
     "",
     "Guardrails:",
@@ -274,6 +278,7 @@ export function buildProofloopProjectManifest(root: string): ProofloopProjectMan
       init: "npx proofloop init --agent auto --live",
       doctor: "npx proofloop doctor --json",
       manifest: "npx proofloop manifest --dense",
+      target: "npx proofloop target --write-runner-plan",
       live: "npx proofloop this-repo --live",
       gate: "npx proofloop gate",
       resume: "npx proofloop resume --dense",
diff --git a/src/targetPlan.ts b/src/targetPlan.ts
index 63ae1be..73a282c 100644
--- a/src/targetPlan.ts
+++ b/src/targetPlan.ts
@@ -1,6 +1,7 @@
 import { existsSync, mkdirSync, readdirSync, readFileSync, writeFileSync } from "node:fs";
 import type { Dirent } from "node:fs";
 import { dirname, join, resolve } from "node:path";
+import { writeProofloopContextReport } from "./contextReport";
 import { buildProofloopLayeredRunnerPlan } from "./layeredPlan";
 import type { ProofloopRunnerPlan, ProofloopRunnerTaskPlan } from "./runner";
 
@@ -52,6 +53,8 @@ export type ProofloopTargetResult = {
   plan: ProofloopTargetPlan;
   planPath: string;
   runnerPlanPath?: string;
+  reportPath?: string;
+  latestReportPath?: string;
 };
 
 export type ProofloopTargetOptions = {
@@ -208,9 +211,15 @@ export async function runProofloopTarget(options: ProofloopTargetOptions): Promi
   try {
     const result = await writeProofloopTargetPlan(options);
     if (options.json) {
-      log(JSON.stringify({ ...result.plan, planPath: result.planPath, runnerPlanPath: result.runnerPlanPath ?? null }, null, 2));
+      log(JSON.stringify({
+        ...result.plan,
+        planPath: result.planPath,
+        runnerPlanPath: result.runnerPlanPath ?? null,
+        reportPath: result.reportPath ?? null,
+        latestReportPath: result.latestReportPath ?? null,
+      }, null, 2));
     } else {
-      log(formatProofloopTargetPlanDense(result.plan, result.planPath, result.runnerPlanPath));
+      log(formatProofloopTargetPlanDense(result.plan, result.planPath, result.runnerPlanPath, result.reportPath));
     }
     return result;
   } catch (error) {
@@ -244,8 +253,16 @@ export async function writeProofloopTargetPlan(options: ProofloopTargetOptions):
     runnerPlanPath = resolve(root, DEFAULT_TARGET_RUNNER_PLAN_PATH);
     writeJson(runnerPlanPath, plan.runnerPlan);
   }
+  const report = writeProofloopContextReport({ root, targetPlan: plan, targetPlanPath: planPath, ...(runnerPlanPath ? { runnerPlanPath } : {}) });
 
-  return { exitCode: 0, plan, planPath, ...(runnerPlanPath ? { runnerPlanPath } : {}) };
+  return {
+    exitCode: 0,
+    plan,
+    planPath,
+    ...(runnerPlanPath ? { runnerPlanPath } : {}),
+    reportPath: report.reportPath,
+    latestReportPath: report.latestPath,
+  };
 }
 
 export function buildProofloopTargetPlan(args: {
@@ -355,7 +372,7 @@ export function classifyBenchmarkFamilies(
   return recommendations.sort((a, b) => b.confidence - a.confidence || a.id.localeCompare(b.id));
 }
 
-export function formatProofloopTargetPlanDense(plan: ProofloopTargetPlan, planPath: string, runnerPlanPath?: string): string {
+export function formatProofloopTargetPlanDense(plan: ProofloopTargetPlan, planPath: string, runnerPlanPath?: string, reportPath?: string): string {
   const lines = [
     "proofloop-target-plan",
     `target=${plan.target.kind}${plan.target.url ? ` url=${plan.target.url}` : ""}${plan.target.packageName ? ` package=${plan.target.packageName}` : ""}`,
@@ -363,6 +380,7 @@ export function formatProofloopTargetPlanDense(plan: ProofloopTargetPlan, planPa
     `officialScoreReady=${String(plan.summary.officialScoreReady)} runnerPlanReady=${String(plan.summary.runnerPlanReady)}`,
     `plan=${planPath}`,
     ...(runnerPlanPath ? [`runnerPlan=${runnerPlanPath}`] : []),
+    ...(reportPath ? [`report=${reportPath}`] : []),
   ];
   for (const rec of plan.recommendations.slice(0, 8)) {
     lines.push(`family=${rec.id} fit=${rec.fit} confidence=${rec.confidence.toFixed(2)} adapter=${rec.adapterStatus} scorer=${rec.officialScoreStatus}`);
diff --git a/tests/agentOsDocs.test.ts b/tests/agentOsDocs.test.ts
new file mode 100644
index 0000000..43a01dc
--- /dev/null
+++ b/tests/agentOsDocs.test.ts
@@ -0,0 +1,20 @@
+import { existsSync, readFileSync } from "node:fs";
+import { join } from "node:path";
+import { describe, expect, it } from "vitest";
+
+const root = process.cwd();
+
+describe("ProofLoop Agent OS docs pack", () => {
+  it("ships the human/agent context doctrine and research notes", () => {
+    const readme = join(root, "docs", "agent-os", "README.md");
+    const research = join(root, "docs", "agent-os", "research.md");
+
+    expect(existsSync(readme)).toBe(true);
+    expect(existsSync(research)).toBe(true);
+
+    expect(readFileSync(readme, "utf8")).toContain("ProofLoop Agent OS Markdown Pack");
+    expect(readFileSync(research, "utf8")).toContain("https://arxiv.org/abs/1803.10122");
+    expect(readFileSync(research, "utf8")).toContain("https://docs.langchain.com/oss/python/langchain/context-engineering");
+    expect(readFileSync(research, "utf8")).toContain("https://www.anthropic.com/engineering/building-effective-agents");
+  });
+});
diff --git a/tests/projectSetup.test.ts b/tests/projectSetup.test.ts
index b4d5648..e464cb4 100644
--- a/tests/projectSetup.test.ts
+++ b/tests/projectSetup.test.ts
@@ -47,7 +47,9 @@ describe("agent-friendly project setup", () => {
 
     const pkg = JSON.parse(readFileSync(join(root, "package.json"), "utf8")) as { scripts: Record<string, string> };
     expect(pkg.scripts["proofloop:init"]).toBe("npx proofloop init --agent auto --live");
+    expect(pkg.scripts["proofloop:target"]).toBe("npx proofloop target --write-runner-plan");
     expect(pkg.scripts["proofloop:charts"]).toBe("npx proofloop charts latest");
+    expect(readFileSync(join(root, "AGENTS.md"), "utf8")).toContain(".proofloop/reports/latest.md");
 
     const manifest = buildProofloopProjectManifest(root);
     expect(manifest.repo.stack).toContain("Vite");
diff --git a/tests/targetPlan.test.ts b/tests/targetPlan.test.ts
index 7dc946d..4ce7cfc 100644
--- a/tests/targetPlan.test.ts
+++ b/tests/targetPlan.test.ts
@@ -59,14 +59,20 @@ describe("proofloop target planner", () => {
 
     const planPath = join(root, ".proofloop", "target", "latest-target-plan.json");
     const runnerPlanPath = join(root, ".proofloop", "runner", "target.plan.json");
+    const reportPath = join(root, ".proofloop", "reports", "latest.md");
     expect(existsSync(planPath)).toBe(true);
     expect(existsSync(runnerPlanPath)).toBe(true);
+    expect(existsSync(reportPath)).toBe(true);
 
     const plan = JSON.parse(readFileSync(planPath, "utf8")) as ProofloopTargetPlan;
+    const report = readFileSync(reportPath, "utf8");
     const accounting = plan.recommendations.find((entry) => entry.id === "bankertoolbench");
     const spreadsheet = plan.recommendations.find((entry) => entry.id === "spreadsheetbench-v1");
 
     expect(plan.target.kind).toBe("codebase");
+    expect(report).toContain("# ProofLoop Context Report");
+    expect(report).toContain("ledger-room");
+    expect(report).toContain("## Not Done / Blocked");
     expect(accounting?.adapterStatus).toBe("configured");
     expect(accounting?.configuredScripts.map((script) => script.name)).toContain("benchmark:accounting");
     expect(accounting?.evidence.join("\n")).toContain("trial balance");
@@ -101,6 +107,8 @@ describe("proofloop target planner", () => {
 
     expect(result.exitCode).toBe(0);
     expect(existsSync(result.planPath)).toBe(true);
+    expect(result.reportPath && existsSync(result.reportPath)).toBe(true);
+    expect(result.latestReportPath && existsSync(result.latestReportPath)).toBe(true);
     expect(result.runnerPlanPath && existsSync(result.runnerPlanPath)).toBe(true);
     expect(result.plan.target.kind).toBe("live-url");
     expect(result.plan.target.httpStatus).toBe(200);
@@ -110,6 +118,7 @@ describe("proofloop target planner", () => {
     expect(result.plan.runnerPlan?.tasks.some((task) => task.id === "target.url-reachable")).toBe(true);
     expect(result.plan.blocked.join("\n")).toContain("no Playwright/Cypress/browser script");
     expect(logs.join("\n")).toContain('"schema": "proofloop-target-plan-v1"');
+    expect(readFileSync(result.latestReportPath!, "utf8")).toContain("## Agent Handoff");
   });
 
   it("can scaffold a Playwright live-smoke adapter for a URL target", async () => {
@@ -143,6 +152,7 @@ describe("proofloop target planner", () => {
     const browser = result.plan.recommendations.find((entry) => entry.id === "live-browser-smoke");
 
     expect(existsSync(specPath)).toBe(true);
+    expect(result.latestReportPath && existsSync(result.latestReportPath)).toBe(true);
     expect(readFileSync(specPath, "utf8")).toContain(url);
     expect(pkg.scripts["proofloop:live-smoke"]).toBe("playwright test proofloop/browser/live-smoke.spec.ts");
     expect(browser?.adapterStatus).toBe("configured");