tella-remove-mistakes
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes video transcripts retrieved from
get_transcriptinSKILL.md. These transcripts are untrusted external data. A transcript containing specific phrases could theoretically trigger mistake-finding rules to cut unintended parts of the video. \n - Ingestion points:
get_transcriptinSKILL.md\n - Boundary markers: Absent. The transcript text is processed directly for linguistic patterns. \n
- Capability inventory:
cut_clip_by_transcriptandupdate_clipinSKILL.md\n - Sanitization: Absent. The skill relies on pattern matching against specific cues (e.g., 'let me start over') without pre-filtering. \n
- Mitigation: The workflow includes a mandatory step to show all findings to the user for approval before performing any cuts, which significantly reduces the risk of malicious or unintended data modification.
Audit Metadata