The Agent Skills Directory

[DATA_EXFILTRATION]: Accesses the sensitive file path ~/.fireflies/env to retrieve the FIREFLIES_API_KEY. While necessary for the skill's functionality, accessing local credential files constitutes data exposure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted meeting transcripts from an external API.
Ingestion points: Meeting sentences and summaries are fetched via curl from https://api.fireflies.ai/graphql (SKILL.md).
Boundary markers: Absent. The transcript text is processed by the agent without delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill uses bash to execute curl, jq, and date commands.
Sanitization: None. The raw text from the transcript is handled without escaping or validation.
[COMMAND_EXECUTION]: Utilizes shell commands (curl, jq, date, bash) to perform network requests, parse JSON data, and convert timestamps.

fireflies-transcript