The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process untrusted data from the internet, creating a surface for injection attacks.\n
Ingestion points: The skill fetches content from arbitrary URLs using tools like agent-browser, WebFetch, and curl, as specified in SKILL.md and the fetcher strategies in references/fetchers/.\n
Boundary markers: The instructions do not define clear boundaries or 'ignore' directives to prevent the agent from following instructions embedded within the fetched web content.\n
Capability inventory: The skill possesses significant capabilities including the ability to Write and Edit files, as well as execute python3, curl, and gh commands, which could be misused if the agent is compromised via injection.\n
Sanitization: Fetched content is passed directly to the analysis and summarization logic without being sanitized or validated for malicious prompts.

content-summarizer