Loading ToolBench
Steganography analysis MCP server — 128 tools, 17 categories. Image steganalysis, JPEG/F5/JSteg detection, audio steganography, video stego, GIF palette analysis, network covert channels, MP3 stego, spread spectrum, BPCS, archive stego, QR code analysis, text steganography, file forensics, document analysis, encoding detection. 100% local, zero API keys required.
Server has 11 visible tools with schemas and descriptions, but quality is inconsistent. All tools have descriptions (good), but many descriptions are verbose (200-400 chars, above the 10-1024 baseline sweet spot of 34-392). Schemas are present but minimal, most tools accept only 1-2 parameters with basic string/number types. No output schemas documented. No error handling guidance. No tool annotations (readOnlyHint/destructiveHint). Three tools (create_*) are destructive but lack confirmation patterns or dry-run support. Parameter descriptions are present but generic. No pagination support despite file-based operations that could return large results. Naming is verb-first and clear (archive_detect, audio_lsb_detect, create_eof_inject), which is good. However, the server claims 128 tools in package.json but only 11 are visible in the provided code, suggests incomplete implementation or missing tool definitions.
Auto-detect steganography in ZIP archives. Checks for slack spaces between entries, prepended data before the first local header, appended data after the end-of-central-directory record, unusual extra fields, and file/archive comments that may conceal hidden data.
ZIP entry structure analysis. Lists all local file headers with their byte offsets, compressed and uncompressed sizes, compression method, CRC-32, flags, and data start offsets. Provides a low-level structural map of the archive.
Auto-detect audio steganography in a WAV file. Runs LSB chi-square, entropy analysis, metadata inspection, and checks for appended data after the WAV data chunk.
PCM sample LSB statistical analysis. Performs chi-square test on LSBs grouped by value pairs to detect LSB replacement steganography in WAV audio.
Auto-detect BPCS (Bit-Plane Complexity Segmentation) steganographic embedding. Computes the complexity map across all bit planes and channels, then checks for an abnormally high ratio of complex blocks — the signature of BPCS embedding which replaces complex bit-plane regions with hidden data.
No output schemas documented. Tools return structured data (findings arrays, analysis results) but LLMs cannot see the expected response shape. This forces LLMs to guess at field names and types, increasing parsing errors.
Destructive tools (create_eof_inject, create_metadata, create_whitespace) lack confirmation or dry-run patterns. Agents can irreversibly modify files without a safety gate. No error recovery guidance if file write fails.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe read-only tools from destructive ones without parsing descriptions. This violates current MCP spec (2026-07-28) which expects tool annotations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Append data after a file's end-of-file marker. The injected data is invisible to most viewers/parsers since they stop reading at the format EOF. Supports any file format — simply appends the given data string to the file's end.
Inject data into a file's metadata fields. For PNG files, inserts a tEXt chunk with the given field name and data. For JPEG files, inserts EXIF APP1 data. The injected metadata persists in the file and is accessible via metadata readers.
Embed data in file whitespace using trailing spaces and tabs on text lines. Each bit of the secret data is encoded as a trailing space (0) or tab (1) appended to lines in the cover text file. The whitespace is invisible in most text editors.
Multi-format decoder supporting Base64, hex, binary, decimal, octal, URL encoding, ROT13, Base32, Morse code, and HTML entities. In auto mode, detects the encoding first and applies the appropriate decoder.
Auto-detect encoding type of an input string. Tests against all known encoding patterns (Base64, hex, binary, morse, URL encoding, HTML entities, etc.) and returns matches sorted by confidence, with attempted decoding for top results.
Character frequency analysis for cryptanalysis. Counts character occurrences, compares to standard English letter frequency (ETAOINSHRDLU), and calculates the Index of Coincidence. IC near 0.0667 suggests English/monoalphabetic, IC near 0.038 suggests random/polyalphabetic.
Package.json claims 128 tools across 17 categories, but only 11 tools are visible in provided source. Either implementation is incomplete or tool definitions are not shown. Cannot verify quality of missing 117 tools.
File-based tools (archive_*, audio_*, image_*) accept file_path as string but provide no validation, error guidance, or handling for missing/unreadable files. Errors will be raw exceptions, not actionable recovery hints.