The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill audits third-party code and documentation, creating a surface for indirect prompt injection.
Ingestion points: The scripts/scrape_docs.py component fetches content from https://docs.typesafe.ai into a local scratch directory for analysis as described in SKILL.md.
Boundary markers: Both SKILL.md and references/spike-protocol.md contain explicit instructions for the agent to treat external content as data and ignore any embedded instructions, credential requests, or endpoint overrides.
Capability inventory: The skill requires file system access to store documentation snapshots and network access for documentation fetching and optional "spike" API calls.
Sanitization: The bundled scrape_docs.py script enforces strict origin checks, prevents path traversal, limits response sizes to 2MB, and validates content types to ensure only text/markdown is processed.
[COMMAND_EXECUTION]: The skill utilizes bundled Python scripts to perform documentation scraping and package validation.
Evidence: SKILL.md references the execution of scripts/scrape_docs.py using the python3 command. The script is an internal part of the skill package and does not pass unvalidated user input to shell commands.

jev-opportunities