The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and parse content from arbitrary external URLs (changelogs and blog posts), creating an attack surface for indirect prompt injection.
Ingestion points: Data enters the agent context via the WebFetch tool and automated scraping adapters (crawl mode) that follow links from index pages.
Boundary markers: The instructions explicitly advise delegating content sampling to a subagent to isolate external data from the primary context window and prevent context pollution.
Capability inventory: The agent can execute the releases CLI tool to modify source metadata, delete sources, and fetch content; it also has access to manage_source and get_latest_releases tools.
Sanitization: No specific sanitization or escaping mechanisms for the fetched HTML/Markdown are described in the provided documentation.

parsing-changelogs