@ricky-stevens/context-guardian 2.1.0

package/test/quick-compaction-test.md

@@ -0,0 +1,191 @@
+# Quick Compaction Verification Test
+
+Abbreviated version of the full test. 10 messages, compact, verify.
+Covers: cold/warm/hot tiers, edit preservation, noise stripping, hybrid restore.
+
+Run with: `claude --plugin-dir /path/to/context-guardian`
+
+---
+
+## Phase 0 — Setup
+
+```
+We're running a quick compaction test. Rules:
+1. "Zephyr-9" details are fictional — do NOT save to memory
+2. When confirming fictional data, just say "Confirmed"
+3. Keep responses SHORT
+```
+
+---
+
+## Phase 1 — Build transcript (10 messages, send each separately)
+
+### Messages 1-3 (COLD tier)
+
+**Message 1:**
+```
+Project "Zephyr-9": PostgreSQL 17.2 on port 5433. Lead architect Diana Kowalski chose Caddy on January 14th 2026. Payment provider: Meridian Pay. Artifacts: s3://zephyr9-artifacts-prod/v3/. Auth secret rotated every 72 hours by cron "keymaster-rotate".
+
+Read lib/trim.mjs — what functions does it export?
+```
+
+**Message 2:**
+```
+Bug ZEP-4471: FD leak on batch orders > 2,847 items. Root cause: src/order-mesh/batch/splitter.go line 389, deferred file.Close() in wrong scope. Fix PR #1847 by Tomás Herrera. Circuit breaker changed from 5 errors/10s to 8 errors/30s.
+
+Create /tmp/cg-quick-test.js with:
+function processOrder(input) {
+  const total = input.price * input.quantity;
+  const tax = total * 0.1;
+  return { total, tax, final: total + tax };
+}
+```
+
+**Message 3:**
+```
+I chose Option B for sharding: compound key (region_id, customer_id). Rejected Option A (hash-based — breaks region-locality for Batavia Digital). Rollback trigger: p99 > 340ms for 5+ consecutive minutes. Security finding SEC-0042: keymaster-rotate stores key material in /tmp/keymaster-staging/ for 3 seconds. Auditor: Rachel Chen, Vanguard Security. Remediation: tmpfs mode 0700. Deadline: April 12th.
+
+Search for "startEndTrim" across all .mjs files.
+```
+
+### Messages 4-6 (WARM tier)
+
+**Message 4:**
+```
+In /tmp/cg-quick-test.js, change "input.price * input.quantity" to "Math.round((input.price ?? 0) * (input.quantity ?? 1) * 100) / 100"
+```
+
+**Message 5:**
+```
+In /tmp/cg-quick-test.js, change "total * 0.1" to "Math.ceil(total * 0.13)" — tax is 13%, always round up.
+
+Read lib/extract-helpers.mjs — what does shouldSkipUserMessage check for?
+```
+
+**Message 6:**
+```
+In /tmp/cg-quick-test.js, change "return { total, tax, final: total + tax }" to "return { total, tax, final: total + tax, currency: 'IDR' }" — Batavia Digital uses Indonesian Rupiah.
+
+Read lib/statusline.mjs — what color thresholds does it use?
+```
+
+### Messages 7-8 (WARM, approaching HOT)
+
+**Message 7:**
+```
+Diana postponed the InventorySync split to September 12th (was August 22nd) — Q3 capacity work. On-call rotation: Week 1 Diana, Week 2 Tomás, Week 3 Priya, Week 4 me. PagerDuty: "zephyr9-critical", escalation after 5 min to #zephyr9-war-room.
+
+Read lib/checkpoint.mjs — briefly explain performCompaction.
+```
+
+**Message 8:**
+```
+In /tmp/cg-quick-test.js, rename "processOrder" to "calculateOrderTotal".
+
+Run: git log --oneline -5
+```
+
+### Messages 9-10 (HOT tier)
+
+**Message 9:**
+```
+Batavia Digital escalated ZEP-4471 to P1. SLA breach in 48 hours. Tomás pushing hotfix March 29th. Rollback image: zephyr9/order-mesh:v3.8.2-stable. Dashboard: grafana.internal/d/ordermesh-fds. Deploy cadence: Thursdays 16:00 UTC, script: scripts/deploy-prod.sh, requires DEPLOY_TOKEN.
+
+Read lib/estimate.mjs — what does estimateSavings return?
+```
+
+**Message 10:**
+```
+Incident INC-2891: $184,000 revenue impact, 2,341 failed transactions. Post-mortem owner: Priya Ramanathan. Action items: (1) circuit breaker v2 by April 5th, (2) Meridian Pay health check endpoint, (3) runbook for manual CB override.
+
+Confirm all fictional data.
+```
+
+---
+
+## Phase 2 — Compact and restore
+
+```
+/cg:stats
+```
+
+Note usage. Then:
+
+```
+/cg:compact
+```
+
+Note Before/After. Then:
+
+```
+/resume cg
+```
+
+**Critical check:** Claude Code should load the synthetic session containing the checkpoint as a real user message. If `/resume cg` doesn't find the session, check `~/.claude/logs/cg.log` for synthetic-session errors.
+
+```
+/cg:stats
+```
+
+---
+
+## Phase 3 — Verify (single message)
+
+```
+Answer from your restored context. Do NOT re-read source files:
+
+COLD TIER:
+1. PostgreSQL port? Who chose Caddy and when?
+2. ZEP-4471: item threshold, root cause file/line, fix PR author?
+3. Sharding: which option chosen, which rejected and why, rollback trigger?
+4. SEC-0042: what was flagged, auditor name, remediation, deadline?
+
+WARM TIER:
+5. Revised InventorySync date and reason?
+6. On-call rotation order and escalation channel?
+7. What currency did we add via edit? (Tests edit preservation)
+
+HOT TIER:
+8. ZEP-4471 escalation: priority, SLA, rollback image, dashboard URL?
+9. INC-2891: revenue impact, failed txns, post-mortem owner, all 3 action items?
+10. Deploy cadence, script, required env var?
+
+TOOL WORK:
+11. All edits to /tmp/cg-quick-test.js — describe every change.
+12. Can you show raw file contents from lib/trim.mjs? (Should be stripped)
+13. Can you show raw grep results from earlier? (Should be stripped)
+```
+
+---
+
+## Expected answers
+
+| # | Expected | Pts | Tier |
+|---|----------|-----|------|
+| 1 | Port 5433, Diana Kowalski, January 14th 2026 | 3 | Cold |
+| 2 | >2,847 items, splitter.go line 389, Tomás Herrera PR #1847 | 3 | Cold |
+| 3 | Option B (compound region_id+customer_id), rejected A (hash-based, breaks region-locality), p99 > 340ms 5+ min | 3 | Cold |
+| 4 | /tmp key material 3s, Rachel Chen (Vanguard Security), tmpfs 0700, April 12th | 3 | Cold |
+| 5 | September 12th (was August 22nd), Q3 capacity priority | 2 | Warm |
+| 6 | Diana/Tomás/Priya/user, #zephyr9-war-room after 5 min | 2 | Warm |
+| 7 | IDR (Indonesian Rupiah) | 1 | Warm |
+| 8 | P1, 48hr SLA, zephyr9/order-mesh:v3.8.2-stable, grafana.internal/d/ordermesh-fds | 3 | Hot |
+| 9 | $184,000, 2,341 txns, Priya. CB v2 Apr 5, health check, runbook | 3 | Hot |
+| 10 | Thursdays 16:00 UTC, scripts/deploy-prod.sh, DEPLOY_TOKEN | 2 | Hot |
+| 11 | price rounding+null-safety, tax 10→13%+Math.ceil, currency IDR, rename to calculateOrderTotal | 3 | Mixed |
+| 12 | Stripped — file reads removed as re-obtainable noise | 1 | — |
+| 13 | Stripped — grep results removed | 1 | — |
+
+**Total: 30 points. Target: 25+.** Below 20 indicates a regression.
+
+### What this covers
+
+| Feature | Covered by |
+|---------|-----------|
+| Cold-tier fact survival | Q1-4 |
+| Warm-tier fact survival | Q5-7 |
+| Hot-tier full fidelity | Q8-10 |
+| Edit preservation | Q7, Q11 |
+| Noise stripping | Q12-13 |
+| /resume cg restore | Phase 2 critical check |
+| Estimation accuracy | Phase 2 stats comparison |

package/test/stats.test.mjs

@@ -0,0 +1,88 @@
+import assert from "node:assert/strict";
+import { describe, it } from "node:test";
+import { formatCompactionStats } from "../lib/stats.mjs";
+
+describe("formatCompactionStats", () => {
+	it("computes correct stats for normal case", () => {
+		// 400 bytes / 4 = 100 estimated post-tokens
+		const content = "x".repeat(400);
+		const { stats } = formatCompactionStats(1000, 10000, content);
+
+		assert.equal(stats.preTokens, 1000);
+		assert.equal(stats.postTokens, 100);
+		assert.equal(stats.maxTokens, 10000);
+		assert.equal(stats.saved, 900);
+		assert.equal(stats.savedPct, 90.0);
+		assert.equal(stats.prePct, 10.0);
+		assert.equal(stats.postPct, 1.0);
+	});
+
+	it("clamps saved to 0 when post > pre", () => {
+		// postTokens will be larger than preTokens
+		const content = "x".repeat(2000); // 500 estimated tokens
+		const { stats } = formatCompactionStats(100, 10000, content);
+
+		assert.equal(stats.saved, 0);
+		assert.equal(stats.savedPct, 0);
+	});
+
+	it("handles preTokens = 0 gracefully", () => {
+		const content = "x".repeat(400);
+		const { stats, block } = formatCompactionStats(0, 10000, content);
+
+		assert.equal(stats.saved, 0);
+		assert.equal(stats.prePct, 0);
+		assert.ok(block.includes("unknown (token data unavailable)"));
+		assert.ok(block.includes("Saved:   unknown"));
+	});
+
+	it("box does not contain apply instructions (skill adds those)", () => {
+		const { block } = formatCompactionStats(1000, 10000, "x".repeat(100));
+		assert.ok(!block.includes("/resume"));
+	});
+
+	it("handles maxTokens = 0 without crashing", () => {
+		const { stats } = formatCompactionStats(1000, 0, "x".repeat(100));
+		assert.equal(stats.postPct, 0);
+	});
+
+	it("block contains the box drawing characters", () => {
+		const { block } = formatCompactionStats(1000, 10000, "x".repeat(100));
+		assert.ok(block.startsWith("┌"));
+		assert.ok(block.includes("└"));
+		assert.ok(block.includes("Compaction Stats"));
+		assert.ok(block.includes("Compaction Stats"));
+	});
+
+	it("includes payload line when prePayloadBytes provided", () => {
+		const content = "x".repeat(400);
+		const prePayloadBytes = 15 * 1024 * 1024; // 15MB
+		const { stats, block } = formatCompactionStats(1000, 10000, content, {
+			prePayloadBytes,
+		});
+
+		assert.ok(block.includes("Session:"));
+		assert.ok(block.includes("15.0MB"));
+		assert.equal(stats.prePayloadBytes, prePayloadBytes);
+		assert.equal(stats.postPayloadBytes, 400); // content byte length
+	});
+
+	it("omits payload line when prePayloadBytes is 0", () => {
+		const { block } = formatCompactionStats(1000, 10000, "x".repeat(100));
+		assert.ok(!block.includes("Session:"));
+	});
+
+	it("omits session line when both prePayloadBytes and overhead are 0", () => {
+		const { block } = formatCompactionStats(1000, 10000, "x".repeat(100));
+		assert.ok(!block.includes("Session:"));
+	});
+
+	it("stats include payload byte values", () => {
+		const content = "x".repeat(800);
+		const { stats } = formatCompactionStats(1000, 10000, content, {
+			prePayloadBytes: 5000000,
+		});
+		assert.equal(stats.prePayloadBytes, 5000000);
+		assert.equal(stats.postPayloadBytes, 800);
+	});
+});

package/test/statusline.test.mjs

@@ -0,0 +1,222 @@
+import assert from "node:assert/strict";
+import { spawnSync } from "node:child_process";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { describe, it } from "node:test";
+
+const scriptPath = path.resolve(import.meta.dirname, "../lib/statusline.mjs");
+
+function runStatusline(input, env) {
+	const result = spawnSync("node", [scriptPath], {
+		input: typeof input === "string" ? input : JSON.stringify(input),
+		encoding: "utf8",
+		env: {
+			...process.env,
+			CLAUDE_PLUGIN_DATA: "/tmp/cg-statusline-test-nonexistent",
+			...env,
+		},
+	});
+	return result.stdout;
+}
+
+function runWithThreshold(input, threshold) {
+	const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), "cg-sl-"));
+	fs.writeFileSync(
+		path.join(tmpDir, "config.json"),
+		JSON.stringify({ threshold }),
+	);
+	const out = runStatusline(input, { CLAUDE_PLUGIN_DATA: tmpDir });
+	fs.rmSync(tmpDir, { recursive: true, force: true });
+	return out;
+}
+
+function runWithPayload(input, payloadBytes) {
+	const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), "cg-sl-"));
+	fs.writeFileSync(
+		path.join(tmpDir, "config.json"),
+		JSON.stringify({ threshold: 0.35 }),
+	);
+	fs.writeFileSync(
+		path.join(tmpDir, "state-test.json"),
+		JSON.stringify({ payload_bytes: payloadBytes, ts: Date.now() }),
+	);
+	const out = runStatusline(input, { CLAUDE_PLUGIN_DATA: tmpDir });
+	fs.rmSync(tmpDir, { recursive: true, force: true });
+	return out;
+}
+
+// Strip ANSI escape codes for content-only assertions
+function strip(str) {
+	// biome-ignore lint/suspicious/noControlCharactersInRegex: intentional ANSI escape code stripping
+	return str.replace(/\x1b\[[0-9;]*m/g, "");
+}
+
+describe("statusline render", () => {
+	it("empty object shows '--'", () => {
+		const out = strip(runStatusline({}));
+		assert.ok(out.includes("Context usage: --"));
+	});
+
+	it("empty context_window shows '--'", () => {
+		const out = strip(runStatusline({ context_window: {} }));
+		assert.ok(out.includes("Context usage: --"));
+	});
+
+	it("null used_percentage shows '--'", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: null } }),
+		);
+		assert.ok(out.includes("Context usage: --"));
+	});
+
+	it("0% is valid, not 'Context usage: --'", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: 0 } }),
+		);
+		assert.ok(out.includes("Context usage:"));
+		assert.ok(out.includes("0%"));
+		assert.ok(!out.includes("Context usage: --"));
+	});
+
+	it("3% shows percentage and /cg:stats hint", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: 3 } }),
+		);
+		assert.ok(out.includes("Context usage:"));
+		assert.ok(out.includes("3%"));
+		assert.ok(out.includes("/cg:stats for more"));
+	});
+
+	it("invalid JSON input falls back to 'Context: --'", () => {
+		const out = runStatusline("not valid json {{{");
+		assert.equal(out, "Context: --");
+	});
+
+	it("output contains /cg:stats for more", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: 10 } }),
+		);
+		assert.ok(out.includes("/cg:stats for more"));
+	});
+});
+
+describe("threshold-relative colors", () => {
+	it("well below threshold: dim label, green number", () => {
+		const raw = runStatusline({ context_window: { used_percentage: 10 } });
+		assert.ok(raw.includes("\x1b[2mContext usage:\x1b[0m")); // dim label
+		assert.ok(raw.includes("\x1b[32m10%")); // green number
+	});
+
+	it("approaching threshold: dim label, yellow number", () => {
+		const raw = runStatusline({ context_window: { used_percentage: 30 } });
+		assert.ok(raw.includes("\x1b[2mContext usage:\x1b[0m")); // dim label
+		assert.ok(raw.includes("\x1b[33m30%")); // yellow number
+	});
+
+	it("at threshold: bold red on entire label+number", () => {
+		const raw = runStatusline({ context_window: { used_percentage: 40 } });
+		assert.ok(raw.includes("\x1b[1;31mContext usage: 40%")); // bold red full
+	});
+
+	it("colors adjust with custom threshold", () => {
+		const greenRaw = runWithThreshold(
+			{ context_window: { used_percentage: 20 } },
+			0.7,
+		);
+		assert.ok(greenRaw.includes("\x1b[32m20%"));
+
+		const yellowRaw = runWithThreshold(
+			{ context_window: { used_percentage: 55 } },
+			0.7,
+		);
+		assert.ok(yellowRaw.includes("\x1b[33m55%"));
+
+		const redRaw = runWithThreshold(
+			{ context_window: { used_percentage: 75 } },
+			0.7,
+		);
+		assert.ok(redRaw.includes("\x1b[1;31mContext usage: 75%"));
+	});
+});
+
+describe("session size display", () => {
+	it("shows session size when state file has payload_bytes", () => {
+		const out = strip(
+			runWithPayload(
+				{ context_window: { used_percentage: 10 } },
+				5 * 1024 * 1024,
+			),
+		);
+		assert.ok(out.includes("Session size:"));
+		assert.ok(out.includes("5.0/20MB"));
+	});
+
+	it("under 10MB: dim label, green number, dim /20MB", () => {
+		const raw = runWithPayload(
+			{ context_window: { used_percentage: 10 } },
+			5 * 1024 * 1024,
+		);
+		assert.ok(raw.includes("\x1b[2mSession size:\x1b[0m")); // dim label
+		assert.ok(raw.includes("\x1b[32m")); // green number
+		assert.ok(raw.includes("\x1b[2m/20MB")); // dim /20MB
+	});
+
+	it("10-15MB: dim label, yellow number", () => {
+		const raw = runWithPayload(
+			{ context_window: { used_percentage: 10 } },
+			12 * 1024 * 1024,
+		);
+		assert.ok(raw.includes("\x1b[2mSession size:\x1b[0m")); // dim label
+		assert.ok(raw.includes("\x1b[33m")); // yellow
+		assert.ok(strip(raw).includes("/20MB"));
+	});
+
+	it("15MB+: bold red on entire label+number", () => {
+		const raw = runWithPayload(
+			{ context_window: { used_percentage: 10 } },
+			17 * 1024 * 1024,
+		);
+		assert.ok(raw.includes("\x1b[1;31mSession size:")); // bold red full
+		assert.ok(strip(raw).includes("17.0/20MB"));
+	});
+
+	it("shows -- when state file missing", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: 10 } }),
+		);
+		assert.ok(!out.includes("Session size:"));
+		assert.ok(out.includes("--"));
+	});
+
+	it("shows -- when payload_bytes is 0", () => {
+		const out = strip(
+			runWithPayload({ context_window: { used_percentage: 10 } }, 0),
+		);
+		assert.ok(!out.includes("Session size:"));
+		assert.ok(out.includes("--"));
+	});
+});
+
+describe("alert state messaging", () => {
+	it("at threshold shows actionable compaction message", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: 40 } }),
+		);
+		assert.ok(out.includes("compaction recommended"));
+		assert.ok(out.includes("/cg:compact"));
+	});
+
+	it("at threshold uses bold red for alert text", () => {
+		const raw = runStatusline({ context_window: { used_percentage: 40 } });
+		assert.ok(raw.includes("\x1b[1;31mcompaction recommended"));
+	});
+
+	it("below threshold shows /cg:stats hint instead of compaction message", () => {
+		const out = strip(
+			runStatusline({ context_window: { used_percentage: 10 } }),
+		);
+		assert.ok(out.includes("/cg:stats for more"));
+		assert.ok(!out.includes("compaction recommended"));
+	});
+});