neozip-mcp 0.9.0-beta → 0.9.5-beta
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +20 -0
- package/README.md +4 -4
- package/dist/archive/extract-entry-buffer.js +4 -1
- package/dist/archive/manifest.js +3 -1
- package/dist/prune/detect.js +3 -1
- package/dist/prune/read-entry.js +14 -7
- package/dist/register/tools.js +2 -2
- package/dist/server.js +1 -1
- package/dist/tools/prune.js +10 -2
- package/dist/tools/test.js +1 -1
- package/docs/OPERATIONS.md +9 -4
- package/docs/examples/CLAUDE.md.example +1 -1
- package/docs/examples/claude/skills/neozip-mcp/SKILL.md +2 -2
- package/docs/examples/neozip-mcp-cursor-rule.mdc +1 -1
- package/docs/installation-guides/README.md +1 -0
- package/package.json +3 -3
package/CHANGELOG.md
CHANGED
|
@@ -4,6 +4,26 @@ Release notes for [neozip-mcp](https://www.npmjs.com/package/neozip-mcp).
|
|
|
4
4
|
|
|
5
5
|
## Unreleased
|
|
6
6
|
|
|
7
|
+
## 0.9.5-beta (2026-11-22)
|
|
8
|
+
|
|
9
|
+
- **neozip-prune 0.5.0.** Pruned archives use the 0.5.0 format: shared parts
|
|
10
|
+
are method 209 (`PruneItem` in `list` / `info`) and each pruned entry names
|
|
11
|
+
its real codec in Extra Field `0x034E`. **Breaking:** archives created by
|
|
12
|
+
0.9.0-beta (neozip-prune 0.1.0) cannot be read; extract them with 0.9.0-beta
|
|
13
|
+
and run `prune_create` again.
|
|
14
|
+
- **PDF pruning disabled.** `prune_analyze` and `prune_create` store PDFs whole
|
|
15
|
+
(`mode: "plain"`, reason `PDF pruning is off`) until PDF stream pruning is
|
|
16
|
+
ready. Office document pruning is unchanged. Pruned archives that other
|
|
17
|
+
NeoZip tools created with PDF parts can still be read and extracted.
|
|
18
|
+
- **SHA-256 kept.** `prune_create` still records each file's SHA-256 (neozip-prune
|
|
19
|
+
now omits it by default), so `extract` / `test` check size, CRC-32 and SHA-256.
|
|
20
|
+
- `read_entry` on a `META-INF/PRUNE/ITEMS/` shared part now explains that it is
|
|
21
|
+
not a user file.
|
|
22
|
+
- **Agent plugin (`plugins/neozip`).** Portable Agent Plugins package in-repo
|
|
23
|
+
(Claude / Cursor / Grok Bot): launches `neozip-mcp@0.9.5-beta` via `npx`, ships
|
|
24
|
+
archive + notarization skills and a Cursor rule. Marketplace index at
|
|
25
|
+
`.cursor-plugin/marketplace.json`. See `plugins/neozip/README.md`.
|
|
26
|
+
|
|
7
27
|
## 0.9.0-beta (2026-11-15)
|
|
8
28
|
|
|
9
29
|
- **Pruned archives (beta).** New `prune_analyze`, `prune_create`, and
|
package/README.md
CHANGED
|
@@ -15,7 +15,7 @@
|
|
|
15
15
|
|
|
16
16
|
## Release notice
|
|
17
17
|
|
|
18
|
-
**Current version: v0.9.
|
|
18
|
+
**Current version: v0.9.5-beta.** See [CHANGELOG.md](CHANGELOG.md) for details.
|
|
19
19
|
|
|
20
20
|
- **Dual-era MCP:** Hosts that still send `initialize` (Claude Code, Cursor) keep working. Hosts that speak MCP 2026-07-28 get the stateless path automatically. One stdio process serves both.
|
|
21
21
|
- **Token Service:** Default remains hosted **testnet** (Base Sepolia). Production (Base Mainnet + Stripe membership) and local (`localhost:14789`) are first-class `neozip-connect` wizard choices.
|
|
@@ -52,7 +52,7 @@ Query portable ZipWiki knowledge packages (OKF concept cards + parsed markdown)
|
|
|
52
52
|
- **wiki_verify** -- CRC-32 / SHA-256 check of every entry
|
|
53
53
|
|
|
54
54
|
### Pruned Archives (beta)
|
|
55
|
-
Store parts shared across Office documents
|
|
55
|
+
Store parts shared across Office documents (fonts, styles, themes, images) once, with the same `neozip-prune` library as `neozip prune` in NeoZip CLI. Documents are rebuilt byte-for-byte on extract. PDFs are stored whole for now (PDF pruning is not enabled yet). `extract`, `test`, `read_entry`, `grep_entries`, `info`, `list` and `zip://` handle pruned archives automatically (only when `META-INF/PRUNE.NZIP` is present).
|
|
56
56
|
- **prune_analyze** -- Report shared parts and prunable bytes (writes nothing)
|
|
57
57
|
- **prune_create** -- Create a verified pruned `.nzip` (`dryRun` to preview)
|
|
58
58
|
- **prune_info** -- Describe a pruned archive (`pruned: false` for regular archives)
|
|
@@ -94,7 +94,7 @@ The [npmjs.com package page](https://www.npmjs.com/package/neozip-mcp) shows **`
|
|
|
94
94
|
|
|
95
95
|
| Command | Problem |
|
|
96
96
|
|---------|---------|
|
|
97
|
-
| `npm i neozip-mcp` |
|
|
97
|
+
| `npm i neozip-mcp` | Missing **`-g`** (CLI bins must be global). Prefer `npm install -g neozip-mcp@0.9.5-beta` or `@latest` / `@beta`. |
|
|
98
98
|
| `npm i neozip-mcp@beta` | Still local-only — `neozip-mcp` / `neozip-connect` won't be on your PATH for MCP hosts |
|
|
99
99
|
|
|
100
100
|
**Use this:**
|
|
@@ -259,7 +259,7 @@ which neozip-mcp
|
|
|
259
259
|
|
|
260
260
|
**`No matching version found`:**
|
|
261
261
|
|
|
262
|
-
-
|
|
262
|
+
- Install with `@latest`, `@beta`, or `@0.9.5-beta` (0.9.5-beta is published as the primary/`latest` version).
|
|
263
263
|
|
|
264
264
|
## Documentation
|
|
265
265
|
|
|
@@ -8,12 +8,15 @@
|
|
|
8
8
|
* Method 208 entries in pruned archives (META-INF/PRUNE.NZIP) are rebuilt by
|
|
9
9
|
* neozip-prune instead; neozipkit refuses that method.
|
|
10
10
|
*/
|
|
11
|
-
import { getPruneContext, isPrunedEntry } from "../prune/detect.js";
|
|
11
|
+
import { PRUNE_ITEM_METHOD, getPruneContext, isPrunedEntry } from "../prune/detect.js";
|
|
12
12
|
import { readPrunedEntryToBuffer } from "../prune/read-entry.js";
|
|
13
13
|
/** In-memory rebuild cap for one pruned entry (read_entry / grep / resources). */
|
|
14
14
|
const PRUNED_ENTRY_MAX_BYTES = 256 * 1024 * 1024;
|
|
15
15
|
export async function extractEntryToBuffer(zip, entry) {
|
|
16
16
|
const prune = getPruneContext(zip);
|
|
17
|
+
if (prune && entry.cmpMethod === PRUNE_ITEM_METHOD) {
|
|
18
|
+
throw new Error(`${entry.filename ?? entry.fileName} is a neozip-prune shared part (META-INF/PRUNE/ITEMS/), not a user file`);
|
|
19
|
+
}
|
|
17
20
|
if (prune && isPrunedEntry(entry)) {
|
|
18
21
|
return readPrunedEntryToBuffer(prune.archivePath, entry.filename ?? entry.fileName, Number(entry.uncompressedSize ?? 0), prune.manifest, PRUNED_ENTRY_MAX_BYTES);
|
|
19
22
|
}
|
package/dist/archive/manifest.js
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { detectTextSafety } from "./detect-text.js";
|
|
2
2
|
import { getEntryEncryptionLabel, isEntryEncrypted } from "./encryption.js";
|
|
3
|
-
import { PRUNE_METHOD } from "../prune/detect.js";
|
|
3
|
+
import { PRUNE_ITEM_METHOD, PRUNE_METHOD } from "../prune/detect.js";
|
|
4
4
|
export function compressionMethodName(cmpMethod) {
|
|
5
5
|
if (cmpMethod === 0)
|
|
6
6
|
return "Stored";
|
|
@@ -10,6 +10,8 @@ export function compressionMethodName(cmpMethod) {
|
|
|
10
10
|
return "Zstd";
|
|
11
11
|
if (cmpMethod === PRUNE_METHOD)
|
|
12
12
|
return "Prune";
|
|
13
|
+
if (cmpMethod === PRUNE_ITEM_METHOD)
|
|
14
|
+
return "PruneItem";
|
|
13
15
|
return `Method ${cmpMethod}`;
|
|
14
16
|
}
|
|
15
17
|
export function buildEntrySummary(entry) {
|
package/dist/prune/detect.js
CHANGED
|
@@ -3,8 +3,10 @@
|
|
|
3
3
|
* carries a META-INF/PRUNE.NZIP manifest that neozip-prune accepts; every other
|
|
4
4
|
* ZIP / NZIP keeps the regular neozipkit path.
|
|
5
5
|
*/
|
|
6
|
-
import { MANIFEST_NAME, PRUNE_METHOD as LIB_PRUNE_METHOD, openPrunedArchive, } from "neozip-prune";
|
|
6
|
+
import { MANIFEST_NAME, PRUNE_METHOD as LIB_PRUNE_METHOD, PRUNE_ITEM_METHOD as LIB_PRUNE_ITEM_METHOD, openPrunedArchive, } from "neozip-prune";
|
|
7
7
|
export const PRUNE_METHOD = LIB_PRUNE_METHOD;
|
|
8
|
+
/** Shared parts under META-INF/PRUNE/ITEMS/ (splice sources, not user files). */
|
|
9
|
+
export const PRUNE_ITEM_METHOD = LIB_PRUNE_ITEM_METHOD;
|
|
8
10
|
const contexts = new WeakMap();
|
|
9
11
|
function entryName(entry) {
|
|
10
12
|
return String(entry?.filename ?? entry?.fileName ?? "");
|
package/dist/prune/read-entry.js
CHANGED
|
@@ -1,34 +1,40 @@
|
|
|
1
1
|
import * as fs from "fs";
|
|
2
|
-
import { restoreNzp1Entry, scanZipLayoutFromFd, } from "neozip-prune";
|
|
2
|
+
import { openPrunedArchive, readPruneMethodExtra, readSha256Extra, restoreNzp1Entry, scanZipLayoutFromFd, } from "neozip-prune";
|
|
3
3
|
/**
|
|
4
|
-
* Rebuild one method-208 entry into memory, checking size
|
|
5
|
-
* against the values
|
|
4
|
+
* Rebuild one method-208 entry into memory, checking size and CRC-32 (and
|
|
5
|
+
* SHA-256 when the archive recorded it) against the values stored at create time.
|
|
6
6
|
*/
|
|
7
7
|
export function readPrunedEntryToBuffer(archivePath, entryName, uncompressedSize, manifest, maxBytes) {
|
|
8
8
|
if (uncompressedSize > maxBytes) {
|
|
9
9
|
throw new Error(`${entryName}: rebuilt size ${uncompressedSize} exceeds limit of ${maxBytes} bytes`);
|
|
10
10
|
}
|
|
11
|
+
const archive = openPrunedArchive(archivePath);
|
|
11
12
|
const fd = fs.openSync(archivePath, "r");
|
|
12
13
|
try {
|
|
14
|
+
const central = new Map(archive.reader.entries().map((e) => [e.name.toUpperCase(), e.centralExtra]));
|
|
13
15
|
const members = new Map();
|
|
14
16
|
for (const m of scanZipLayoutFromFd(fd, fs.fstatSync(fd).size).members) {
|
|
15
|
-
|
|
17
|
+
const key = m.name.toUpperCase();
|
|
18
|
+
const extra = central.get(key);
|
|
19
|
+
members.set(key, {
|
|
16
20
|
name: m.name,
|
|
17
21
|
dataStart: m.dataStart,
|
|
18
22
|
compressedSize: m.compressedSize,
|
|
19
23
|
method: m.method,
|
|
24
|
+
realMethod: extra ? readPruneMethodExtra(extra) : null,
|
|
20
25
|
});
|
|
21
26
|
}
|
|
22
|
-
const
|
|
27
|
+
const key = entryName.toUpperCase();
|
|
28
|
+
const loc = members.get(key);
|
|
23
29
|
if (!loc)
|
|
24
30
|
throw new Error(`Entry not found: ${entryName}`);
|
|
25
|
-
const
|
|
31
|
+
const extra = central.get(key);
|
|
26
32
|
const chunks = [];
|
|
27
33
|
restoreNzp1Entry({
|
|
28
34
|
archiveFd: fd,
|
|
29
35
|
entry: loc,
|
|
30
36
|
uncompressedSize,
|
|
31
|
-
declaredSha256:
|
|
37
|
+
declaredSha256: extra ? readSha256Extra(extra) : null,
|
|
32
38
|
manifest,
|
|
33
39
|
members,
|
|
34
40
|
onChunk: (chunk) => chunks.push(Buffer.from(chunk)),
|
|
@@ -37,6 +43,7 @@ export function readPrunedEntryToBuffer(archivePath, entryName, uncompressedSize
|
|
|
37
43
|
}
|
|
38
44
|
finally {
|
|
39
45
|
fs.closeSync(fd);
|
|
46
|
+
archive.close();
|
|
40
47
|
}
|
|
41
48
|
}
|
|
42
49
|
//# sourceMappingURL=read-entry.js.map
|
package/dist/register/tools.js
CHANGED
|
@@ -62,8 +62,8 @@ export function registerTools(deps) {
|
|
|
62
62
|
registerTool(deps, "wiki_origin", "Resolve the original document for a parsed entry (Extra Field 0x014F origin). With fetch: true, download and verify CRC-32 (requires NEOZIP_MCP_WIKI_ALLOW_ORIGIN_FETCH=true); optional dest writes the verified file inside the sandbox.", wikiOriginParameters, (args) => executeWikiOrigin(args, wikiDeps));
|
|
63
63
|
registerTool(deps, "wiki_verify", "Verify every entry of a .zipwiki package (CRC-32, SHA-256 when present) and flag paths that do not belong in a ZipWiki package.", wikiVerifyParameters, (args) => executeWikiVerify(args, wikiDeps));
|
|
64
64
|
const pruneDeps = { sandbox, resourceLimiter, config };
|
|
65
|
-
registerTool(deps, "prune_analyze", "Beta. Dry run of neozip-prune: list the parts of Office documents (.docx/.xlsx/.pptx, other ZIP containers) and
|
|
66
|
-
registerTool(deps, "prune_create", "Beta. Create a pruned .nzip from files/directories: parts shared across documents are stored once under META-INF/PRUNE/ITEMS/ and each document is rebuilt byte-for-byte on extract (verified by default). Best for collections of Office documents
|
|
65
|
+
registerTool(deps, "prune_analyze", "Beta. Dry run of neozip-prune: list the parts of Office documents (.docx/.xlsx/.pptx, other ZIP containers) and report the shared parts (fonts, styles, themes, images) and bytes a pruned archive would save. PDFs and other files are reported as plain (PDF pruning is not enabled yet). Read-only.", pruneAnalyzeParameters, (args) => executePruneAnalyze(args, pruneDeps));
|
|
66
|
+
registerTool(deps, "prune_create", "Beta. Create a pruned .nzip from files/directories: parts shared across documents are stored once under META-INF/PRUNE/ITEMS/ and each document is rebuilt byte-for-byte on extract (verified by default). Best for collections of Office documents from one source; PDFs and other files are stored whole. Method 208 entries need neozip tools to extract. options.dryRun analyzes only.", pruneCreateParameters, (args) => executePruneCreate(args, pruneDeps));
|
|
67
67
|
registerTool(deps, "prune_info", "Beta. Describe a pruned archive (META-INF/PRUNE.NZIP): format version, shared parts, and per-entry mode/sizes. Returns pruned: false for regular archives. extract/test/read_entry/grep_entries/info/list already handle pruned archives.", pruneInfoParameters, (args) => executePruneInfo(args, pruneDeps));
|
|
68
68
|
registerTool(deps, "mint", "Mint a ZIP file's merkle root as an NFT on the blockchain (Base network). Requires wallet configuration.", mintParameters, (args) => executeMint(args, auth, resourceLimiter, config, config.network));
|
|
69
69
|
registerTool(deps, "verify", "Verify a tokenized ZIP file's authenticity by checking its TOKEN.NZIP metadata against the blockchain. Pass options.password if payload entries are encrypted.", verifyParameters, (args) => executeVerify(args, sandbox, resourceLimiter, config));
|
package/dist/server.js
CHANGED
|
@@ -9,7 +9,7 @@ import { registerResources } from "./register/resources.js";
|
|
|
9
9
|
export function createServer(config) {
|
|
10
10
|
const server = new FastMCP({
|
|
11
11
|
name: "NeoZip MCP Server",
|
|
12
|
-
version: "0.9.
|
|
12
|
+
version: "0.9.5-beta",
|
|
13
13
|
});
|
|
14
14
|
const sandbox = new PathSandbox(config.sandboxPaths);
|
|
15
15
|
const capabilities = new CapabilityScope(config.capabilities);
|
package/dist/tools/prune.js
CHANGED
|
@@ -22,6 +22,8 @@ function toList(input) {
|
|
|
22
22
|
return Array.isArray(input) ? input : [input];
|
|
23
23
|
}
|
|
24
24
|
const TOP_PARTS = 20;
|
|
25
|
+
/** PDF part pruning (and its experimental Flate pack) is off until it is production-ready; PDFs are stored whole. */
|
|
26
|
+
const PRUNE_OPTIONS = { pdf: false };
|
|
25
27
|
function summarizeAnalysis(report) {
|
|
26
28
|
return {
|
|
27
29
|
files: report.files,
|
|
@@ -38,7 +40,7 @@ export const pruneAnalyzeParameters = z.object({
|
|
|
38
40
|
export function executePruneAnalyze(args, deps) {
|
|
39
41
|
return runPrune("prune_analyze", deps, async () => {
|
|
40
42
|
const files = await expandPruneInputs(toList(args.input), deps.sandbox, deps.resourceLimiter);
|
|
41
|
-
return summarizeAnalysis(analyzeFiles(files));
|
|
43
|
+
return summarizeAnalysis(analyzeFiles(files, PRUNE_OPTIONS));
|
|
42
44
|
});
|
|
43
45
|
}
|
|
44
46
|
// --- prune_create ------------------------------------------------------------
|
|
@@ -73,15 +75,21 @@ export function executePruneCreate(args, deps) {
|
|
|
73
75
|
const opts = args.options ?? {};
|
|
74
76
|
const files = await expandPruneInputs(toList(args.input), deps.sandbox, deps.resourceLimiter, outputPath);
|
|
75
77
|
if (opts.dryRun) {
|
|
76
|
-
return {
|
|
78
|
+
return {
|
|
79
|
+
dryRun: true,
|
|
80
|
+
output: outputPath,
|
|
81
|
+
...summarizeAnalysis(analyzeFiles(files, PRUNE_OPTIONS)),
|
|
82
|
+
};
|
|
77
83
|
}
|
|
78
84
|
if (fs.existsSync(outputPath) && !opts.overwrite) {
|
|
79
85
|
throw new Error(`Output already exists: ${outputPath} (set options.overwrite to replace it)`);
|
|
80
86
|
}
|
|
81
87
|
await fs.promises.mkdir(path.dirname(outputPath), { recursive: true });
|
|
82
88
|
const result = await createPrunedArchive(outputPath, files, {
|
|
89
|
+
...PRUNE_OPTIONS,
|
|
83
90
|
level: opts.level ?? 7,
|
|
84
91
|
verify: opts.verify ?? true,
|
|
92
|
+
sha256: true,
|
|
85
93
|
});
|
|
86
94
|
return {
|
|
87
95
|
output: result.path,
|
package/dist/tools/test.js
CHANGED
|
@@ -86,7 +86,7 @@ export async function executeTest(args, sandbox, resourceLimiter, config) {
|
|
|
86
86
|
const lines = [];
|
|
87
87
|
lines.push(`Testing: ${archivePath}`);
|
|
88
88
|
if (prune) {
|
|
89
|
-
lines.push("Pruned archive (META-INF/PRUNE.NZIP): method 208 entries rebuilt and checked (size, CRC-32, SHA-256)");
|
|
89
|
+
lines.push("Pruned archive (META-INF/PRUNE.NZIP): method 208 entries rebuilt and checked (size, CRC-32, and SHA-256 when recorded)");
|
|
90
90
|
}
|
|
91
91
|
lines.push("");
|
|
92
92
|
for (const r of results) {
|
package/docs/OPERATIONS.md
CHANGED
|
@@ -301,7 +301,7 @@ Archive path, file size on disk, entry count, uncompressed/compressed totals, co
|
|
|
301
301
|
|
|
302
302
|
### `test`
|
|
303
303
|
|
|
304
|
-
Test the integrity of a ZIP archive by verifying each file entry. In pruned archives, method 208 entries are rebuilt and checked against their recorded size
|
|
304
|
+
Test the integrity of a ZIP archive by verifying each file entry. In pruned archives, method 208 entries are rebuilt and checked against their recorded size and CRC-32 (and SHA-256 when recorded).
|
|
305
305
|
|
|
306
306
|
**Parameters**
|
|
307
307
|
|
|
@@ -628,17 +628,22 @@ Check every entry (CRC-32, SHA-256 when present) and flag paths that do not belo
|
|
|
628
628
|
|
|
629
629
|
## Pruned Archives (beta)
|
|
630
630
|
|
|
631
|
-
A pruned archive is a `.nzip` built by **neozip-prune** (the same library as `neozip prune` in NeoZip CLI). Parts that repeat across documents, such as fonts, styles, themes and images in Office files (`.docx`, `.xlsx`, `.pptx` and other ZIP containers)
|
|
631
|
+
A pruned archive is a `.nzip` built by **neozip-prune** (the same library as `neozip prune` in NeoZip CLI). Parts that repeat across documents, such as fonts, styles, themes and images in Office files (`.docx`, `.xlsx`, `.pptx` and other ZIP containers), are stored once under `META-INF/PRUNE/ITEMS/` (compression method **209**). Each document is stored as a small skeleton entry (compression method **208**) that is rebuilt byte-for-byte on extraction. Files with no shared parts are stored as plain zstd entries.
|
|
632
|
+
|
|
633
|
+
**PDFs are stored whole.** PDF stream pruning is not enabled in neozip-mcp yet; `prune_analyze` and `prune_create` report PDFs as `plain` with reason `PDF pruning is off`. Pruned archives that other NeoZip tools created with PDF parts can still be read, tested and extracted.
|
|
634
|
+
|
|
635
|
+
Archives written by neozip-mcp 0.9.0-beta (neozip-prune 0.1.0) use an older format that neozip-prune 0.5.0 cannot read; extract them with 0.9.0-beta and run `prune_create` again. Format reference: `neozip-prune/docs/NEOZIP_PRUNE_FORMAT.md` in the neozip-libs repo.
|
|
632
636
|
|
|
633
637
|
- An archive is treated as pruned **only** when it contains a valid `META-INF/PRUNE.NZIP` manifest. Regular `.zip` / `.nzip` archives (and `.zipwiki` packages) are not affected.
|
|
634
638
|
- The existing tools handle pruned archives automatically:
|
|
635
|
-
- **extract** rebuilds method 208 entries (size
|
|
639
|
+
- **extract** rebuilds method 208 entries (size and CRC-32 checked, plus SHA-256 when recorded) and honors `overwrite`. The prune sidecars (`META-INF/PRUNE.NZIP`, `META-INF/PRUNE/**`) are not written to the output.
|
|
636
640
|
- **test** rebuilds and checks every method 208 entry; a damaged shared part reports `FAIL: integrity_failed`.
|
|
637
641
|
- **read_entry**, **grep_entries** and `zip://` resources rebuild the entry in memory (text entries only, as for regular archives).
|
|
638
642
|
- **list** / **info** show method `Prune` and `archiveKind: "pruned"`; JSON output adds a `prune` block. No Merkle root is reported for pruned archives.
|
|
639
643
|
- **stamp** refuses pruned archives (timestamp/tokenize of pruned archives is not supported yet). `compress` has no prune option; use `prune_create`.
|
|
640
644
|
- Other unzip tools list method 208 entries but cannot extract them. Share pruned archives only with people who use NeoZip tools.
|
|
641
|
-
-
|
|
645
|
+
- `prune_create` records each file's SHA-256, so extract and test check size, CRC-32 and SHA-256. Archives from other tools may omit SHA-256; those are checked by size and CRC-32.
|
|
646
|
+
- Best fit: collections of Office documents from one source (shared templates, fonts, branding). Run `prune_analyze` (or `prune_create` with `dryRun`) first to see the savings.
|
|
642
647
|
|
|
643
648
|
### `prune_analyze`
|
|
644
649
|
|
|
@@ -5,7 +5,7 @@ This project uses **neozip-mcp** for ZIP and `.nzip` archive operations via MCP
|
|
|
5
5
|
## Agent behavior
|
|
6
6
|
|
|
7
7
|
- Use **neozip-mcp** MCP tools for archive work — do not read or search for neozip-mcp source unless this repo is the neozip-mcp development repository.
|
|
8
|
-
- **Archives** (create, read, extract, search, inspect): skills under `.claude/skills/neozip-mcp/` — tools `compress`, `extract`, `list`, `info`, `test`, `read_entry`, `search_entries`, `grep_entries`; `.zipwiki` packages via `wiki_open`, `wiki_search`, `wiki_query`, `wiki_read`; pruned archives of Office
|
|
8
|
+
- **Archives** (create, read, extract, search, inspect): skills under `.claude/skills/neozip-mcp/` — tools `compress`, `extract`, `list`, `info`, `test`, `read_entry`, `search_entries`, `grep_entries`; `.zipwiki` packages via `wiki_open`, `wiki_search`, `wiki_query`, `wiki_read`; pruned archives of Office document collections (beta) via `prune_analyze`, `prune_create`, `prune_info`.
|
|
9
9
|
- **Notarization / recipient encryption** (tokenize, timestamp, verify, encrypt for email): skills under `.claude/skills/neozip-notarization/` — call `connect_status` first; if not ready, tell the user to run `neozip-connect` and start a new session.
|
|
10
10
|
- Use **absolute paths** under the MCP sandbox (`NEOZIP_MCP_SANDBOX_PATHS`, typically `${CLAUDE_PROJECT_DIR:-.}` and `${HOME}`).
|
|
11
11
|
- Surface MCP tool errors (`Error: …`) verbatim to the user.
|
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
name: neozip-mcp
|
|
3
3
|
description: >-
|
|
4
4
|
Create, read, extract, search, and inspect ZIP and .nzip archives, and query
|
|
5
|
-
.zipwiki knowledge packages, and create pruned archives of Office
|
|
5
|
+
.zipwiki knowledge packages, and create pruned archives of Office document
|
|
6
6
|
collections, via neozip-mcp MCP tools. Use when the user mentions zip, nzip,
|
|
7
7
|
zipwiki, prune, archives, compress, extract, unzip, or NeoZip; or
|
|
8
8
|
when inspecting archive contents (listing, reading one file, searching names or
|
|
@@ -44,7 +44,7 @@ These tools only query. To create or update a package, use ZipWiki (`zipwiki pac
|
|
|
44
44
|
|
|
45
45
|
## Pruned archives (beta)
|
|
46
46
|
|
|
47
|
-
For a collection of Office documents (`.docx`, `.xlsx`, `.pptx`)
|
|
47
|
+
For a collection of Office documents (`.docx`, `.xlsx`, `.pptx`) from one source, `prune_create` stores the parts they share (fonts, styles, themes, images) once in a `.nzip`. PDFs can be included but are stored whole — do not suggest pruning to save space on PDFs. Run `prune_analyze` (or `prune_create` with `options.dryRun`) first to see the savings; `prune_info` describes a pruned archive.
|
|
48
48
|
|
|
49
49
|
`extract`, `test`, `read_entry`, `grep_entries`, `info` and `list` handle pruned archives automatically — they are detected only by a `META-INF/PRUNE.NZIP` manifest, so regular archives are unaffected. Other unzip tools cannot extract the pruned (method 208) entries: tell the user before sharing a pruned archive with someone who does not use NeoZip tools. `stamp` refuses pruned archives.
|
|
50
50
|
|
|
@@ -18,7 +18,7 @@ For create, read, extract, search, or inspect ZIP/.nzip archives, use **neozip-m
|
|
|
18
18
|
|
|
19
19
|
For `.zipwiki` knowledge packages: `wiki_open` → `wiki_search` / `wiki_query` → `wiki_read` (OKF card first, then parsed) → `wiki_origin` / `wiki_extract` only when needed.
|
|
20
20
|
|
|
21
|
-
Pruned archives (beta): `prune_analyze` → `prune_create` (`.nzip`) for collections of Office documents
|
|
21
|
+
Pruned archives (beta): `prune_analyze` → `prune_create` (`.nzip`) for collections of Office documents that share fonts, styles and images (PDFs are stored whole); `prune_info` to inspect. `extract`, `test`, `read_entry` and `grep_entries` handle pruned archives automatically; other unzip tools cannot extract them.
|
|
22
22
|
|
|
23
23
|
Do **not** read, search, or assume access to neozip-mcp source code. End users install the published npm package only.
|
|
24
24
|
|
|
@@ -9,6 +9,7 @@ Step-by-step setup for **neozip-mcp** (**beta**) in common MCP host applications
|
|
|
9
9
|
| **Claude Code** (CLI, user/local scope) | [INSTALL_CLAUDE_CODE.md](./INSTALL_CLAUDE_CODE.md) | `~/.claude.json`, `~/.claude/skills/` |
|
|
10
10
|
| **Claude Code workspace** (shared project `.mcp.json`) | [INSTALL_CLAUDE_WORKSPACE.md](./INSTALL_CLAUDE_WORKSPACE.md) | `.mcp.json`, `.claude/skills/`, `CLAUDE.md` |
|
|
11
11
|
| **Cursor** | [package README](../README.md#cursor-setup) | `~/.cursor/mcp.json` or `.cursor/mcp.json` |
|
|
12
|
+
| **Agent plugin** (Claude / Cursor / Grok Bot) | [plugins/neozip/README.md](../../plugins/neozip/README.md) | `plugins/neozip/` in this repo |
|
|
12
13
|
|
|
13
14
|
## Before you configure MCP
|
|
14
15
|
|
package/package.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "neozip-mcp",
|
|
3
|
-
"version": "0.9.
|
|
4
|
-
"release_date": "2026-11-
|
|
3
|
+
"version": "0.9.5-beta",
|
|
4
|
+
"release_date": "2026-11-22",
|
|
5
5
|
"description": "NeoZip MCP server for ZIP archives, blockchain notarization, and AI agent integration (Beta)",
|
|
6
6
|
"main": "dist/index.js",
|
|
7
7
|
"bin": {
|
|
@@ -54,7 +54,7 @@
|
|
|
54
54
|
"ethers": "^6.0.0",
|
|
55
55
|
"minimatch": "^10.0.3",
|
|
56
56
|
"neozip-blockchain": "^1.6.3",
|
|
57
|
-
"neozip-prune": "^0.
|
|
57
|
+
"neozip-prune": "^0.5.0",
|
|
58
58
|
"neozip-wiki": "^0.1.0",
|
|
59
59
|
"neozipkit": "^1.6.3",
|
|
60
60
|
"zod": "^3.23.0"
|