feat(memory-ingest): honor the per-remote deny/read-only trust policy (#2392)

Transcript ingest now respects the same trust store as code import — the gate
existed only in gstack-gbrain-sync's runCodeImport, so memory-ingest happily
ingested transcripts from deny-listed repos. preparePages filters prepared
transcript pages through ONE batch policy lookup (new 'get --batch' verb on
bin/gstack-gbrain-repo-policy — the script owns URL normalization; the client
adds repoPolicyTierBatch, one spawn for all distinct remotes, so large corpora
never pay a 10s-timeout subprocess per remote).

Outcomes match code-import semantics: read-only → clean skip
(skipped_policy_readonly), deny → counted refusal (skipped_policy_deny),
corrupted/unreadable store → HARD ERROR before any write (state, staging,
egress receipt, and import all untouched) with the recovery command named —
policy corruption must never read as successful ingestion. Artifacts are
never policy-filtered (their git_remote is a project slug, not a remote).

Fixes #2392.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Garry Tan
2026-08-17 10:39:24 -07:00
co-authored by Claude Fable 5
parent 40e4a53f74
commit 2494276742
6 changed files with 632 additions and 2 deletions
+50 -1
View File
@@ -7,6 +7,13 @@
# if no URL is passed. Exits 0 with one of: read-write, read-only,
# deny, unset.
#
# gstack-gbrain-repo-policy get --batch
# Read remote URLs from stdin (one per line); print one tier per line
# in input order: read-write, read-only, deny, or none (no entry / no
# store). A corrupt store is a hard error (exit 2), NEVER quarantined:
# batch callers are unattended ingest gates that must fail closed
# rather than bypass a set policy.
#
# gstack-gbrain-repo-policy set <remote-url> <read-write|read-only|deny>
# Persist a tier for the given remote. Exits 0 on success.
#
@@ -161,8 +168,50 @@ ensure_file() {
fi
}
# get --batch — bulk lookup for ingest gates. One URL per stdin line, one
# tier per stdout line, input order preserved. Reuses normalize() (the same
# code path single `get` uses) per line. Prints `none` where single `get`
# prints `unset` — batch consumers (lib/gbrain-repo-policy-client.ts) speak
# the RepoPolicyTierValue vocabulary directly.
#
# Corruption polarity differs from single `get` ON PURPOSE: interactive
# `get` quarantines a corrupt store and starts fresh because /setup-gbrain
# re-asks the user; a batch caller is an unattended ingest gate with nobody
# to re-ask, so silently quarantining would BYPASS a set deny policy. Batch
# fails hard (exit 2) instead and names the recovery path.
cmd_get_batch() {
require_jq
if [ ! -f "$POLICY_FILE" ]; then
# No store = no policy was ever set. Every URL is `none`; don't create
# the file just for a read (matches cmd_list).
while IFS= read -r url || [ -n "$url" ]; do
printf 'none\n'
done
return 0
fi
if ! jq empty "$POLICY_FILE" 2>/dev/null; then
die "policy store $POLICY_FILE is corrupt (invalid JSON) — refusing batch read. Inspect with: gstack-gbrain-repo-policy list; re-run /setup-gbrain to rebuild the store."
fi
# Valid JSON from here, so ensure_file only performs the legacy
# allow → read-write migration (never the quarantine branch).
ensure_file
local url key
while IFS= read -r url || [ -n "$url" ]; do
key=$(normalize "$url")
if [ -z "$key" ]; then
printf 'none\n'
continue
fi
jq -r --arg key "$key" '.[$key] // "none"' "$POLICY_FILE"
done
}
cmd_get() {
local url="${1:-}"
if [ "$url" = "--batch" ]; then
cmd_get_batch
return 0
fi
if [ -z "$url" ]; then
url=$(git remote get-url origin 2>/dev/null || true)
if [ -z "$url" ]; then
@@ -221,7 +270,7 @@ case "${1:-}" in
set) shift; cmd_set "$@" ;;
list) shift; cmd_list "$@" ;;
normalize) shift; cmd_normalize "$@" ;;
--help|-h|help) sed -n '2,47p' "$0" | sed 's/^# \{0,1\}//' ;;
--help|-h|help) sed -n '2,54p' "$0" | sed 's/^# \{0,1\}//' ;;
"") die "usage: gstack-gbrain-repo-policy {get|set|list|normalize|--help}" ;;
*) die "unknown subcommand: $1" ;;
esac
+133 -1
View File
@@ -68,6 +68,7 @@ import {
import { execGbrainText, spawnGbrainAsync } from "../lib/gbrain-exec";
import { writeReceipt } from "../lib/egress-receipt";
import { checkOwnedStagingDir, STAGING_MARKER } from "../lib/staging-guard";
import { hasRepoPolicyStore, repoPolicyTierBatch } from "../lib/gbrain-repo-policy-client";
// ── Types ──────────────────────────────────────────────────────────────────
@@ -150,6 +151,14 @@ interface BulkResult {
skipped_secret: number;
skipped_dedup: number;
skipped_unattributed: number;
/**
* #2392: transcripts skipped because their git remote's trust tier in
* ~/.gstack/gbrain-repo-policy.json is `read-only` (search allowed, page
* writes never — and transcript ingest writes pages).
*/
skipped_policy_readonly: number;
/** #2392: transcripts skipped because their remote's trust tier is `deny`. */
skipped_policy_deny: number;
failed: number;
duration_ms: number;
partial_pages: number;
@@ -914,6 +923,14 @@ interface PreparedPage {
/** Carry-through fields for state recording on success. */
page_slug: string;
partial: boolean;
/** Memory type — the per-remote policy filter (#2392) applies to transcripts only. */
type: MemoryType;
/**
* Canonical git remote ("host/org/repo") for transcript pages; undefined
* for artifacts (whose PageRecord.git_remote is a project slug, not a
* remote — artifacts are never policy-filtered).
*/
git_remote?: string;
}
interface StagingResult {
@@ -1210,8 +1227,16 @@ function preparePages(
skippedSecret: number;
skippedDedup: number;
skippedUnattributed: number;
skippedPolicyReadonly: number;
skippedPolicyDeny: number;
parseFailed: number;
partialPages: number;
/**
* #2392: set when the per-remote policy store EXISTS but could not be
* read (corrupt file, spawn failure). The caller must abort before any
* writes — proceeding would bypass a possibly-set deny policy.
*/
policyError?: string;
} {
const prepared: PreparedPage[] = [];
let skippedSecret = 0;
@@ -1280,16 +1305,78 @@ function preparePages(
rendered_body: renderPageBody(page),
page_slug: page.slug,
partial: page.partial ?? false,
type,
// Only transcripts carry a real remote; buildArtifactPage's git_remote
// is a project slug, and artifacts are never policy-filtered (#2392).
git_remote: type === "transcript" ? page.git_remote : undefined,
});
}
// #2392: per-remote trust policy for transcript pages — the same store the
// code-import gate honors (bin/gstack-gbrain-sync.ts). One batch spawn for
// all distinct remotes in the run; no store on disk → zero policy work.
// Runs AFTER the loop because preparePages accumulates fully in memory (no
// writes happen until the caller stages), so filtering here is still
// strictly before any write.
let finalPrepared = prepared;
let skippedPolicyReadonly = 0;
let skippedPolicyDeny = 0;
let policyError: string | undefined;
if (hasRepoPolicyStore()) {
const remotes = [
...new Set(
prepared
.filter((p) => p.type === "transcript" && p.git_remote)
.map((p) => p.git_remote as string),
),
];
if (remotes.length > 0) {
const verdicts = repoPolicyTierBatch(remotes);
// The store EXISTS (checked above), so an unreadable/spawn-failed
// result is a HARD ERROR — match the fail-closed polarity of
// gstack-gbrain-sync's code-import gate: never bypass a set policy.
const broken = remotes.find((r) => {
const v = verdicts.get(r);
return !v || v.error !== undefined;
});
if (broken) {
const kind = verdicts.get(broken)?.error === "spawn-failed"
? "the policy helper could not be spawned (bash missing from PATH?)"
: "the policy store could not be read (corrupt file?)";
policyError =
`repo policy store exists but ${kind} — refusing transcript ingest rather than ` +
`bypassing a possibly-set deny policy. Inspect with: gstack-gbrain-repo-policy list; ` +
`re-run /setup-gbrain if the store is corrupt.`;
} else {
finalPrepared = prepared.filter((p) => {
if (p.type !== "transcript" || !p.git_remote) return true;
const tier = verdicts.get(p.git_remote)?.tier ?? "none";
if (tier === "read-only") {
// Honoring an explicit user setting (search allowed, page writes
// never) — transcript ingest writes pages, so skip.
skippedPolicyReadonly++;
return false;
}
if (tier === "deny") {
skippedPolicyDeny++;
return false;
}
return true; // read-write, or none (no policy set for this remote)
});
}
}
}
return {
prepared,
prepared: finalPrepared,
skippedSecret,
skippedDedup,
skippedUnattributed,
skippedPolicyReadonly,
skippedPolicyDeny,
parseFailed,
partialPages,
policyError,
};
}
@@ -1636,6 +1723,25 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
let written = 0;
let failed = 0;
// #2392 HARD ERROR: the policy store exists but could not be consulted.
// Abort before ANY write — state recording, staging, gbrain import — so a
// corrupt store can never silently bypass a set deny/read-only policy.
if (prep.policyError) {
console.error(`[memory-ingest] ERR: ${prep.policyError}`);
return {
written: 0,
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed: prep.parseFailed + prep.prepared.length,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
system_error: prep.policyError,
};
}
if (args.noWrite) {
// --no-write: skip the gbrain import call but still record state for
// prepared pages (treat them as ingested for dedup purposes). Matches
@@ -1664,6 +1770,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed: prep.parseFailed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1680,6 +1788,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed: prep.parseFailed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1695,6 +1805,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed: prep.parseFailed + prep.prepared.length,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1842,6 +1954,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1881,6 +1995,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1917,6 +2033,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1935,6 +2053,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -1963,6 +2083,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -2015,6 +2137,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -2094,6 +2218,8 @@ async function ingestPass(args: CliArgs): Promise<BulkResult> {
skipped_secret: prep.skippedSecret,
skipped_dedup: prep.skippedDedup,
skipped_unattributed: prep.skippedUnattributed,
skipped_policy_readonly: prep.skippedPolicyReadonly,
skipped_policy_deny: prep.skippedPolicyDeny,
failed: failed + prep.parseFailed,
duration_ms: Date.now() - t0,
partial_pages: prep.partialPages,
@@ -2140,6 +2266,12 @@ function printBulkResult(r: BulkResult, args: CliArgs): void {
console.log(` skipped (dedup): ${r.skipped_dedup}`);
console.log(` skipped (secret-scan): ${r.skipped_secret}`);
console.log(` skipped (unattrib): ${r.skipped_unattributed}`);
if (r.skipped_policy_readonly > 0) {
console.log(` skipped (policy read-only): ${r.skipped_policy_readonly} (remote tier is read-only; transcript ingest writes pages)`);
}
if (r.skipped_policy_deny > 0) {
console.log(` skipped (policy deny): ${r.skipped_policy_deny} (change with: gstack-gbrain-repo-policy set <remote> read-write)`);
}
console.log(` failed: ${r.failed}`);
console.log(` duration: ${(r.duration_ms / 1000).toFixed(1)}s`);
if (args.benchmark) {
+68
View File
@@ -82,3 +82,71 @@ export function repoPolicyTier(url: string | null, env: NodeJS.ProcessEnv = proc
if (tier === "unset") return { tier: "none" };
return { tier: "none", error: "unreadable" }; // unexpected output — a read failure, not a tier
}
/**
* Bulk trust-tier lookup via `gstack-gbrain-repo-policy get --batch` — ONE
* spawn total for the whole url list (memory-ingest checks every distinct
* transcript remote in a run; per-url spawns would fork N bash+jq processes).
*
* The deduped url list goes to the script's stdin, one per line; the script
* answers one tier per line in input order (`none` where single `get` says
* `unset`). Fast paths mirror repoPolicyTier: no store on disk → every url
* is `{ tier: "none" }` with no subprocess.
*
* Failure classification matches repoPolicyTier (spawn ENOENT →
* `spawn-failed`, everything else → `unreadable`), applied to EVERY url: a
* malformed, incomplete, or timed-out batch (wrong line count, unknown tier
* token, non-zero exit) maps every url to `{ tier: "none", error:
* "unreadable" }`. POLARITY IS STILL THE CALLER'S — this client only reads
* and classifies.
*/
export function repoPolicyTierBatch(
urls: string[],
env: NodeJS.ProcessEnv = process.env,
): Map<string, RepoPolicyResult> {
const out = new Map<string, RepoPolicyResult>();
const distinct = [...new Set(urls)];
if (distinct.length === 0) return out;
if (!hasRepoPolicyStore(env)) {
for (const u of distinct) out.set(u, { tier: "none" });
return out;
}
const allWithError = (error: "unreadable" | "spawn-failed"): Map<string, RepoPolicyResult> => {
for (const u of distinct) out.set(u, { tier: "none", error });
return out;
};
// Same win32 bash-wrapping as repoPolicyTier: the script is
// `#!/usr/bin/env bash`, which win32 can't exec directly.
const [cmd, args]: [string, string[]] =
process.platform === "win32"
? ["bash", [POLICY_SCRIPT, "get", "--batch"]]
: [POLICY_SCRIPT, ["get", "--batch"]];
const res = spawnSync(cmd, args, {
encoding: "utf-8",
timeout: 10_000,
input: distinct.join("\n") + "\n",
env: { ...env } as NodeJS.ProcessEnv,
});
if (res.error) {
const code = (res.error as NodeJS.ErrnoException).code;
return allWithError(code === "ENOENT" ? "spawn-failed" : "unreadable");
}
if (res.status !== 0) return allWithError("unreadable");
const lines = (res.stdout || "").replace(/\n$/, "").split("\n");
if (lines.length !== distinct.length) return allWithError("unreadable");
const parsed: RepoPolicyResult[] = [];
for (const raw of lines) {
const tier = raw.trim();
if (tier === "deny" || tier === "read-only" || tier === "read-write") {
parsed.push({ tier });
} else if (tier === "none") {
parsed.push({ tier: "none" });
} else {
// Unknown token anywhere poisons the whole batch — a partially-garbled
// response can't be trusted line-by-line (the ordering itself may be off).
return allWithError("unreadable");
}
}
for (let i = 0; i < distinct.length; i++) out.set(distinct[i], parsed[i]);
return out;
}
+19
View File
@@ -35,6 +35,25 @@ happens after you say yes.
- **Repos under a `deny` trust policy** (set in `/setup-gbrain` Step 6)
are skipped — neither code nor transcripts from those repos ingest.
## Per-remote trust policy (deny / read-only)
Transcript ingest respects the same per-remote trust store as code import
(`~/.gstack/gbrain-repo-policy.json`, managed by
`gstack-gbrain-repo-policy`). Each transcript's git remote is checked
against the store before anything is written:
- **deny** — the transcript is skipped (reported as `skipped (policy deny)`).
- **read-only** — skipped too: read-only means "search allowed, page
writes never", and transcript ingest writes pages (reported as
`skipped (policy read-only)`).
- **read-write, or no entry** — ingests normally.
- **Corrupted or unreadable store** — ingestion aborts before any writes
rather than bypassing a set policy. Inspect the store with
`gstack-gbrain-repo-policy list`; re-run `/setup-gbrain` if it's corrupt.
Artifacts (learnings, plans, retros, etc.) are never policy-filtered — the
policy is keyed by git remote, which artifacts don't have.
## What gets scanned for secrets
The cross-machine secret boundary is `gstack-brain-sync` (the git push
+152
View File
@@ -0,0 +1,152 @@
/**
* lib/gbrain-repo-policy-client — batch trust-tier lookup (#2392).
*
* Covers the `get --batch` verb of bin/gstack-gbrain-repo-policy (bash level:
* multiple urls in, per-line verdicts out, input order preserved) and the
* TypeScript client `repoPolicyTierBatch` (ONE spawn, dedup, fast paths, and
* the whole-batch `unreadable` classification on a corrupt store).
*
* Each test uses a temp GSTACK_HOME so nothing leaks into the user's real
* ~/.gstack. The client is exercised against the REAL bash script — the
* script owns URL normalization, so stub stores are seeded through its own
* `set` verb.
*/
import { describe, test, expect, beforeEach, afterEach } from "bun:test";
import * as fs from "fs";
import * as path from "path";
import * as os from "os";
import { spawnSync } from "child_process";
import { repoPolicyTierBatch } from "../lib/gbrain-repo-policy-client";
const ROOT = path.resolve(import.meta.dir, "..");
const BIN = path.join(ROOT, "bin", "gstack-gbrain-repo-policy");
let tmpHome: string;
function env(): NodeJS.ProcessEnv {
return { ...process.env, GSTACK_HOME: tmpHome };
}
function run(args: string[], input?: string) {
const res = spawnSync(BIN, args, { env: env(), encoding: "utf-8", input });
return {
stdout: res.stdout || "",
stderr: res.stderr || "",
status: res.status ?? -1,
};
}
function policyFile(): string {
return path.join(tmpHome, "gbrain-repo-policy.json");
}
beforeEach(() => {
tmpHome = fs.mkdtempSync(path.join(os.tmpdir(), "gbrain-policy-client-"));
});
afterEach(() => {
fs.rmSync(tmpHome, { recursive: true, force: true });
});
describe("bin/gstack-gbrain-repo-policy get --batch (bash level)", () => {
test("multiple urls in, per-line verdicts out, input order preserved", () => {
expect(run(["set", "https://github.com/foo/bar.git", "deny"]).status).toBe(0);
expect(run(["set", "git@github.com:baz/qux.git", "read-only"]).status).toBe(0);
expect(run(["set", "https://github.com/rw/repo", "read-write"]).status).toBe(0);
const r = run(
["get", "--batch"],
// Mixed URL forms — the script's normalize() collapses them to the
// stored keys. `nope/never` has no entry → none.
"git@github.com:foo/bar.git\nhttps://github.com/nope/never\nhttps://github.com/baz/qux\nhttps://github.com/rw/repo.git\n",
);
expect(r.status).toBe(0);
expect(r.stdout).toBe("deny\nnone\nread-only\nread-write\n");
});
test("no store on disk: every line is none, and no file is created", () => {
const r = run(["get", "--batch"], "https://github.com/a/a\nhttps://github.com/b/b\n");
expect(r.status).toBe(0);
expect(r.stdout).toBe("none\nnone\n");
expect(fs.existsSync(policyFile())).toBe(false);
});
test("corrupt store: hard error (exit 2), NOT quarantined, names recovery", () => {
fs.writeFileSync(policyFile(), "not valid json{", { mode: 0o600 });
const r = run(["get", "--batch"], "https://github.com/foo/bar\n");
expect(r.status).toBe(2);
expect(r.stderr).toContain("corrupt");
expect(r.stderr).toContain("gstack-gbrain-repo-policy list");
// Unlike interactive `get`, batch must never quarantine-and-proceed —
// that would bypass a set deny policy on an unattended ingest run.
expect(fs.readFileSync(policyFile(), "utf-8")).toBe("not valid json{");
expect(
fs.readdirSync(tmpHome).find((f) => f.includes(".corrupt-")),
).toBeUndefined();
});
test("legacy allow entries migrate to read-write on batch read", () => {
fs.writeFileSync(
policyFile(),
JSON.stringify({ "github.com/foo/bar": "allow" }),
{ mode: 0o600 },
);
const r = run(["get", "--batch"], "https://github.com/foo/bar\n");
expect(r.status).toBe(0);
expect(r.stdout).toBe("read-write\n");
});
});
describe("repoPolicyTierBatch (TypeScript client)", () => {
test("maps each input url to its verdict, dedup included", () => {
expect(run(["set", "https://github.com/foo/bar", "deny"]).status).toBe(0);
expect(run(["set", "https://github.com/baz/qux", "read-only"]).status).toBe(0);
const verdicts = repoPolicyTierBatch(
[
"github.com/foo/bar", // canonical form, as memory-ingest passes it
"github.com/baz/qux",
"github.com/nope/never",
"github.com/foo/bar", // duplicate — dedup keeps ONE map entry
],
env(),
);
expect(verdicts.size).toBe(3);
expect(verdicts.get("github.com/foo/bar")).toEqual({ tier: "deny" });
expect(verdicts.get("github.com/baz/qux")).toEqual({ tier: "read-only" });
expect(verdicts.get("github.com/nope/never")).toEqual({ tier: "none" });
});
test("no store on disk: every url is tier none with no error (fast path)", () => {
const verdicts = repoPolicyTierBatch(["github.com/a/a", "github.com/b/b"], env());
expect(verdicts.get("github.com/a/a")).toEqual({ tier: "none" });
expect(verdicts.get("github.com/b/b")).toEqual({ tier: "none" });
expect(fs.existsSync(policyFile())).toBe(false);
});
test("empty url list returns an empty map without spawning", () => {
const verdicts = repoPolicyTierBatch([], env());
expect(verdicts.size).toBe(0);
});
test("corrupt store: EVERY url maps to { tier: none, error: unreadable }", () => {
fs.writeFileSync(policyFile(), "not valid json{", { mode: 0o600 });
const verdicts = repoPolicyTierBatch(["github.com/foo/bar", "github.com/baz/qux"], env());
expect(verdicts.get("github.com/foo/bar")).toEqual({ tier: "none", error: "unreadable" });
expect(verdicts.get("github.com/baz/qux")).toEqual({ tier: "none", error: "unreadable" });
});
test("store unreadable on disk (chmod 000): whole batch classified unreadable", () => {
if (process.platform === "win32" || process.getuid?.() === 0) return; // chmod semantics differ
expect(run(["set", "https://github.com/foo/bar", "deny"]).status).toBe(0);
fs.chmodSync(policyFile(), 0o000);
try {
const verdicts = repoPolicyTierBatch(["github.com/foo/bar"], env());
expect(verdicts.get("github.com/foo/bar")).toEqual({ tier: "none", error: "unreadable" });
} finally {
fs.chmodSync(policyFile(), 0o600);
}
});
});
+210
View File
@@ -939,3 +939,213 @@ describe("#2394: probe applies the same attribution gate as prepare", () => {
rmSync(home, { recursive: true, force: true });
});
});
// ── #2392: transcript ingest honors the per-remote trust policy ─────────────
//
// The same store the code-import gate honors (bin/gstack-gbrain-sync.ts):
// tier `deny` and `read-only` transcripts are skipped with their own counters;
// a store that EXISTS but can't be read is a hard error before any writes
// (never a silent bypass of a set policy); no store at all = zero policy work.
// The policy store is seeded through the REAL bin/gstack-gbrain-repo-policy
// script (its `set` verb owns the file schema + URL normalization).
describe("#2392: transcript ingest honors per-remote trust policy", () => {
const POLICY_BIN = join(import.meta.dir, "..", "bin", "gstack-gbrain-repo-policy");
/** Attributable temp git repo whose origin points at `remoteUrl`. */
function makeRepoWithRemote(home: string, name: string, remoteUrl: string): string {
const repo = join(home, "work", name);
mkdirSync(repo, { recursive: true });
spawnSync("git", ["-C", repo, "init", "-q"], { encoding: "utf-8" });
spawnSync("git", ["-C", repo, "remote", "add", "origin", remoteUrl], { encoding: "utf-8" });
return repo;
}
function writeSessionForRepo(home: string, projectName: string, sessionId: string, cwd: string): void {
const record = JSON.stringify({
type: "user",
message: { role: "user", content: `hello from ${sessionId}` },
timestamp: new Date().toISOString(),
cwd,
});
writeClaudeCodeSession(home, projectName, sessionId, record + "\n");
}
function setPolicy(gstackHome: string, url: string, tier: string): void {
const r = spawnSync(POLICY_BIN, ["set", url, tier], {
encoding: "utf-8",
env: { ...process.env, GSTACK_HOME: gstackHome },
});
expect(r.status).toBe(0);
}
function stateSessions(gstackHome: string): string[] {
const statePath = join(gstackHome, ".transcript-ingest-state.json");
if (!existsSync(statePath)) return [];
return Object.keys(JSON.parse(readFileSync(statePath, "utf-8")).sessions || {});
}
it("(a) deny remote's transcript is skipped and counted as skipped_policy_deny", () => {
const home = makeTestHome();
const gstackHome = join(home, ".gstack");
mkdirSync(gstackHome, { recursive: true });
const { binDir, logFile } = installFakeGbrain(home);
const denyCwd = makeRepoWithRemote(home, "denied", "https://github.com/denyme/denied.git");
const okCwd = makeRepoWithRemote(home, "allowed", "https://github.com/okorg/okrepo.git");
writeSessionForRepo(home, "work-denied", "denysess1", denyCwd);
writeSessionForRepo(home, "work-allowed", "oksess1", okCwd);
setPolicy(gstackHome, "https://github.com/denyme/denied.git", "deny");
const r = runScript(["--bulk", "--quiet"], {
HOME: home,
GSTACK_HOME: gstackHome,
PATH: `${binDir}:${process.env.PATH || ""}`,
});
expect(r.exitCode).toBe(0);
expect(r.stdout).toMatch(/written:\s+1/);
expect(r.stdout).toMatch(/skipped \(policy deny\):\s+1/);
expect(r.stdout).not.toMatch(/skipped \(policy read-only\)/);
// Only the allowed session was imported + state-recorded.
expect(existsSync(logFile)).toBe(true);
const sessions = stateSessions(gstackHome);
expect(sessions.length).toBe(1);
expect(sessions[0]).toContain("oksess1");
rmSync(home, { recursive: true, force: true });
});
it("(b) read-only remote's transcript is skipped and counted as skipped_policy_readonly", () => {
const home = makeTestHome();
const gstackHome = join(home, ".gstack");
mkdirSync(gstackHome, { recursive: true });
const { binDir } = installFakeGbrain(home);
const roCwd = makeRepoWithRemote(home, "readonly", "https://github.com/roorg/rorepo.git");
const okCwd = makeRepoWithRemote(home, "allowed", "https://github.com/okorg/okrepo.git");
writeSessionForRepo(home, "work-readonly", "rosess1", roCwd);
writeSessionForRepo(home, "work-allowed", "oksess1", okCwd);
setPolicy(gstackHome, "https://github.com/roorg/rorepo.git", "read-only");
const r = runScript(["--bulk", "--quiet"], {
HOME: home,
GSTACK_HOME: gstackHome,
PATH: `${binDir}:${process.env.PATH || ""}`,
});
expect(r.exitCode).toBe(0);
expect(r.stdout).toMatch(/written:\s+1/);
expect(r.stdout).toMatch(/skipped \(policy read-only\):\s+1/);
const sessions = stateSessions(gstackHome);
expect(sessions.length).toBe(1);
expect(sessions[0]).toContain("oksess1");
rmSync(home, { recursive: true, force: true });
});
it("(c) read-write remote's transcript is ingested (reaches gbrain import)", () => {
const home = makeTestHome();
const gstackHome = join(home, ".gstack");
mkdirSync(gstackHome, { recursive: true });
const { binDir, logFile } = installFakeGbrain(home);
const rwCwd = makeRepoWithRemote(home, "readwrite", "https://github.com/rworg/rwrepo.git");
writeSessionForRepo(home, "work-readwrite", "rwsess1", rwCwd);
setPolicy(gstackHome, "https://github.com/rworg/rwrepo.git", "read-write");
const r = runScript(["--bulk", "--quiet"], {
HOME: home,
GSTACK_HOME: gstackHome,
PATH: `${binDir}:${process.env.PATH || ""}`,
});
expect(r.exitCode).toBe(0);
expect(r.stdout).toMatch(/written:\s+1/);
expect(r.stdout).not.toMatch(/skipped \(policy/);
// gbrain import ran exactly once — the page reached the import stage.
const calls = readFileSync(logFile, "utf-8").trim().split("\n").filter(Boolean);
expect(calls.length).toBe(1);
expect(stateSessions(gstackHome).length).toBe(1);
rmSync(home, { recursive: true, force: true });
});
it("(d) corrupted store: hard error before any writes, message names recovery", () => {
const home = makeTestHome();
const gstackHome = join(home, ".gstack");
mkdirSync(gstackHome, { recursive: true });
const { binDir, logFile } = installFakeGbrain(home);
const cwd = makeRepoWithRemote(home, "somerepo", "https://github.com/some/repo.git");
writeSessionForRepo(home, "work-somerepo", "somesess1", cwd);
// Corrupt store — the batch verb refuses (exit 2), the client classifies
// `unreadable`, and ingest must abort rather than bypass a set policy.
writeFileSync(join(gstackHome, "gbrain-repo-policy.json"), "not valid json{", "utf-8");
const r = runScript(["--bulk", "--quiet"], {
HOME: home,
GSTACK_HOME: gstackHome,
PATH: `${binDir}:${process.env.PATH || ""}`,
});
expect(r.exitCode).toBe(1);
expect(r.stderr).toMatch(/\[memory-ingest\] ERR:.*repo policy store exists/);
expect(r.stderr).toContain("gstack-gbrain-repo-policy list");
expect(r.stderr).toContain("/setup-gbrain");
// Nothing written: no gbrain import call, no state file, store untouched.
expect(existsSync(logFile)).toBe(false);
expect(stateSessions(gstackHome).length).toBe(0);
expect(readFileSync(join(gstackHome, "gbrain-repo-policy.json"), "utf-8")).toBe("not valid json{");
rmSync(home, { recursive: true, force: true });
});
it("(e) no store at all: no policy filtering, transcript ingests normally", () => {
const home = makeTestHome();
const gstackHome = join(home, ".gstack");
mkdirSync(gstackHome, { recursive: true });
const { binDir } = installFakeGbrain(home);
const cwd = makeRepoWithRemote(home, "freerepo", "https://github.com/free/repo.git");
writeSessionForRepo(home, "work-freerepo", "freesess1", cwd);
const r = runScript(["--bulk", "--quiet"], {
HOME: home,
GSTACK_HOME: gstackHome,
PATH: `${binDir}:${process.env.PATH || ""}`,
});
expect(r.exitCode).toBe(0);
expect(r.stdout).toMatch(/written:\s+1/);
expect(r.stdout).not.toMatch(/skipped \(policy/);
expect(stateSessions(gstackHome).length).toBe(1);
rmSync(home, { recursive: true, force: true });
});
it("artifacts are never policy-filtered, even when their project's remote is denied", () => {
const home = makeTestHome();
const gstackHome = join(home, ".gstack");
mkdirSync(join(gstackHome, "projects", "denyme-denied"), { recursive: true });
const { binDir } = installFakeGbrain(home);
// A learning artifact under a project slug matching a denied remote —
// the policy is keyed by git remote, which artifacts don't have.
writeFileSync(join(gstackHome, "projects", "denyme-denied", "learnings.jsonl"), '{"key":"a","insight":"b"}\n');
setPolicy(gstackHome, "https://github.com/denyme/denied.git", "deny");
const r = runScript(["--bulk", "--quiet"], {
HOME: home,
GSTACK_HOME: gstackHome,
PATH: `${binDir}:${process.env.PATH || ""}`,
});
expect(r.exitCode).toBe(0);
expect(r.stdout).toMatch(/written:\s+1/);
expect(r.stdout).not.toMatch(/skipped \(policy/);
rmSync(home, { recursive: true, force: true });
});
});