The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs several third-party tools from unverified GitHub repositories using go install, including github.com/003random/getJS, github.com/BishopFox/jsluice/cmd/jsluice, and github.com/denandz/sourcemapper. These sources are not recognized as trusted or well-known organizations.
[EXTERNAL_DOWNLOADS]: The skill uses npm install -g to download and install the js-beautify and webcrack packages from the public NPM registry at runtime.
[EXTERNAL_DOWNLOADS]: In Step 3, the skill performs a pip install -r requirements.txt on a local directory, which triggers the installation of unverified Python dependencies.
[COMMAND_EXECUTION]: The skill executes multiple complex shell pipelines, including an xargs command that spawns subshells to perform parallel downloads via curl and file hashing via md5sum.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted data from external targets.
Ingestion points: Remote JavaScript files are identified via katana, gau, or httpx and downloaded to the local filesystem (SKILL.md, Steps 1 & 2).
Boundary markers: Absent; the content of the downloaded files is processed directly by various analytical tools without encapsulation or warning markers.
Capability inventory: The skill possesses capabilities for subprocess execution (executing Go/Python tools), file system writes (writing loot to directories), and network access (downloading target JS files).
Sanitization: There is no evidence of validation or sanitization of the downloaded JavaScript content before it is passed to analytical tools such as LinkFinder, SecretFinder, or jsluice.

js-analysis