linkedin-post-generator
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill inserts user-provided data directly into shell commands. Placeholders like {url}, {search_term}, and {channel_name} in the script execution paths (e.g., scrape_linkedin_posts.py and slack_search.py) are not sanitized, which could lead to arbitrary command execution if a user provides input with shell special characters.
- [INDIRECT_PROMPT_INJECTION]: The skill pulls data from external, potentially untrusted sources like Slack and meeting transcripts and includes them in the generation process without safety boundaries. 1. Ingestion points: The source-material folder (populated by Fireflies, Slack, and Drive). 2. Boundary markers: None; the skill reads entire files and passes them to the model. 3. Capability inventory: Shell command execution (via script orchestration) and file access to user-local configuration. 4. Sanitization: No input filtering or prompt escaping is applied to the ingested content.
Audit Metadata