fireflies-core-workflow-a
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external transcript data from Fireflies.ai, which represents an ingestion surface for untrusted content.
- Ingestion points: Meeting content is ingested via the Fireflies GraphQL API
transcriptquery (defined in SKILL.md). - Boundary markers: The instructions mandate a "redaction and retention boundary" and require field classification before content usage (SKILL.md).
- Capability inventory: The skill utilizes standard filesystem tools (Read, Glob, Grep, Write, Edit) for code and documentation management.
- Sanitization: The instructions explicitly require the removal of sensitive identifiers and media URLs unless authorized, and advise against treating live captions as finalized data.
- [EXTERNAL_DOWNLOADS]: The skill references official documentation and API endpoints from
fireflies.ai. As this is a well-known service, these references are considered standard for the intended integration and do not involve execution of untrusted remote code. - [CREDENTIALS_UNSAFE]: The skill follows security best practices by instructing the agent to use an approved secret manager for the
FIREFLIES_API_KEYand strictly forbidding the inclusion of the key in logs, URLs, or output.
Audit Metadata