The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of meeting transcripts, summaries, and speaker data from the Fireflies.ai API into the agent context, which represents an attack surface for indirect prompt injection if the meeting content contains malicious instructions.
Ingestion points: The FirefliesClient implementation in SKILL.md includes methods like getTranscripts and getTranscript that fetch external data from https://api.fireflies.ai/graphql.
Boundary markers: The provided templates do not include explicit delimiters or system instructions to ignore instructions potentially embedded within the transcript text.
Capability inventory: The skill configuration in SKILL.md allows the Read, Write, and Edit tools.
Sanitization: While Step 4 demonstrates response validation using the Zod library, this focuses on schema integrity rather than sanitizing natural language content for injection patterns.

fireflies-sdk-patterns