ig-reel
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data and external reel transcripts to generate personalized content and infer a voice profile.
- Ingestion points: User raw ideas and transcripts of existing reels mentioned in
SKILL.md. - Boundary markers: Absent; there are no specific delimiters to prevent embedded instructions in transcripts from affecting agent behavior.
- Capability inventory: Local file read/write access to the
~/.claude/instagram/directory and execution of provided Python scripts (hookscore.py,beats.py). - Sanitization: Absent; the skill does not explicitly validate or filter the input text for malicious prompts.
- Note: This represents a standard risk surface for agentic processing of user content and is considered low risk for this specific use case.
- [SAFE]: The skill uses local resources and provided scripts to perform its functions. No attempts at credential harvesting, data exfiltration to non-whitelisted domains, or unauthorized system modifications were detected. The file operations are limited to a specific configuration and log directory suitable for the agent's context.
Audit Metadata