The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous bash commands and a bash script to be executed by the agent or user. These commands utilize standard utilities like curl, grep, sed, and for loops to process web data and local file system paths in /tmp.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading JavaScript files and Source Maps from external target domains for analysis. It also references the well-known Wayback Machine (web.archive.org) for historical data retrieval.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites (JavaScript files). This represents a surface for indirect prompt injection if the extracted content contains instructions intended to influence the agent's behavior during the analysis phase.
Ingestion points: JavaScript file content fetched via curl in SKILL.md and references/js-extract-patterns.md.
Boundary markers: Absent; there are no specific delimiters used to isolate processed data from the instructions.
Capability inventory: Full shell execution capabilities including curl, grep, and sed are used throughout the scripts.
Sanitization: Absent; the skill relies on regular expression patterns to extract specific strings without performing validation or escaping on the content.

js-api-extract