registry / @bastra-recall/daemon / 0.7.6

@bastra-recall/daemon@0.7.6

MCP server exposing recall + load_memory over a markdown vault.

Static Scan Results

scanned 14d ago · by rust-scanner

Static analysis completed at 65.0% confidence. No malicious behavior was detected; 10 low-signal pattern(s) were surfaced and cleared.

Static reason

No blocking static signals were detected.

Decision evidence

public snapshot

Behavioral surface

Source

ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStringsUrlStrings

ManifestNo manifest risk signals triggered.

scanned 67 file(s), 660 KB of source, external domains: 127.0.0.1, api.github.com, github.com, ollama.com, your.host

Source & flagged code

3 flagged

dist/statusline-session.jsView file

Lines 1-33javascript

1/**

2 * Per-session statusline feed paths.

3 *

4 * Claude Code does NOT send a session identifier to MCP servers

5 * (anthropics/claude-code#41836), so the forwarder cannot know which CC

6 * session it belongs to from the protocol. But the forwarder, the

7 * prompt-hook, and the statusline subprocess of ONE session all share the

8 * same `claude` ancestor process. We walk the parent chain up to that

9 * `claude` process and use its PID to namespace the feed file — so

10 * concurrent sessions never clobber each other's counters.

11 *

12 * Feed layout: ~/.bastra/statusline/<claude-session-pid>.json

13 */

14import { execFileSync } from "node:child_process";

15import { mkdirSync, readdirSync, unlinkSync, writeFileSync } from "node:fs";

High

Child Process

Package source references child process execution.

dist/statusline-session.jsView on unpkg · L13

16import * as os from "node:os";

17import * as path from "node:path";

18export const STATUSLINE_DIR = path.join(os.homedir(), ".bastra", "statusline");

19/**

20 * Shared, session-independent vault size, written by the daemon on every index

21 * change and read by every session's statusline segment.

22 *

23 * Why separate from the per-session feed: the forwarder only refreshes a

24 * session's `vault_size` on that session's own tool calls, so an idle session

25 * (no recalls) keeps showing a stale count after another session saves a

26 * memory. The daemon owns the vault and sees every add/remove, so it publishes

27 * the live count here; the segment prefers it over the per-session value.

28 *

29 * Not matched by reapStaleFeeds (which only deletes `<pid>.json`), so it

30 * survives feed reaping.

31 */

32export const SHARED_VAULT_FILE = path.join(STATUSLINE_DIR, "vault.json");

33export function writeSharedVaultSize(size) {

dist/usage-sidecar.jsView file

Lines 56-96javascript

56 entry[e.kind] += 1;

57 const tsField = `last_${e.kind}_at`;

58 // Events arrive roughly ordered, but a merge after compaction rotation may

59 // replay older lines — keep the max, not the last seen.

dist/mcp-forwarder.jsView file

Lines 21-61javascript

21 *

22 * Bootstrap:

23 * 1. GET /health probieren. 200 → Daemon läuft, weiter.

24 * 2. Sonst: detached

Source & flagged code

3 flagged · loading source

dist/statusline-session.jsView file

Static Scan Results

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code