edgespeak-translate
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcript data in JSON or SRT format which presents an attack surface for indirect prompt injection attacks.
- Ingestion points: The agent reads source segments from external transcript files as part of the translation process (SKILL.md Step 3). The verification script also reads these files to check invariants.
- Boundary markers: The instructions do not specify any delimiters or safety markers to isolate transcript text or instruct the agent to ignore embedded commands.
- Capability inventory: The agent is directed to write translation results to the file system (SKILL.md Step 6) and is encouraged to parallelize tasks by dispatching subagents (SKILL.md Step 4).
- Sanitization: The check-translation.mjs script verifies structural data like segment counts and timestamps but does not analyze or sanitize the translated text content for malicious prompt injection payloads.
Audit Metadata