The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data during its research phases, creating a surface for indirect prompt injection.
Ingestion points: The skill fetches data from web searches and competitor content in Step 2: Research Existing Content and Step 3: Competitor Research (SKILL.md).
Boundary markers: There are no explicit instructions to use delimiters or to ignore instructions embedded within the ingested external data.
Capability inventory: The skill uses the Agent tool to run subagents, performs web searches, accesses platform-specific analytics tools, and reads/writes to the local file system (SKILL.md).
Sanitization: No sanitization or filtering of external content is specified before processing or document generation.
[EXTERNAL_DOWNLOADS]: The skill provides references to several external industry blogs and marketing resources in references/youtube-community-strategy.md for strategic context.
Details: The references include links to domains such as vidiq.com, buffer.com, influencermarketinghub.com, geeky-gadgets.com, socioblend.com, air.io, socialmediatoday.com, sproutsocial.com, and hootsuite.com. These are recognized as standard industry sources for social media strategy.
[DATA_EXFILTRATION]: The skill is designed to access sensitive user metrics and platform data.
Evidence: The skill instructions specify gathering "performance metrics", "YouTube Analytics", "email platform data", and "platform analytics" in SKILL.md. This data is intended to be synthesized into a research document saved at a location specified by the user or orchestrator.

research