youtube-full
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes content from third-party YouTube transcripts. An attacker could embed malicious instructions within a video's captions to influence the agent's behavior during summarization or analysis.
- Ingestion points: The skill fetches external transcript data via the
get_transcriptoperation described inSKILL.md. - Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the fetched transcript data as untrusted or to ignore embedded instructions.
- Capability inventory: The agent uses network access to fetch data and then synthesizes the information for the user, creating a chain where injected instructions could be executed.
- Sanitization: There is no evidence of filtering or sanitization of the transcript text before it is returned to the agent's context.
- [DATA_EXFILTRATION]: The skill makes outbound HTTPS requests to
transcriptapi.com. This is the documented and intended communication required for the skill to function and does not involve the transmission of sensitive local files or credentials.
Audit Metadata