hindsight
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
hindsightCLI to manage memory operations and uses shell utilities likebasenameandgitto determine repository-specific memory banks. It also provides a Python script to audit the local configuration of the memory plugin. - [DATA_EXFILTRATION]: The skill is designed to capture and persist user messages, interaction history, and system configurations. It communicates with an API endpoint (
api.hs.delo.sh) located on the author's infrastructure to store and synchronize this data. - [EXTERNAL_DOWNLOADS]: Documentation within the skill references official installers and SDKs available via
vectorize.ioand standard package registries such as npm and PyPI for the initial setup and management of the memory engine. - [PROMPT_INJECTION]: The skill ingests untrusted data from user messages and project metadata to extract facts. While it utilizes boundary markers like response citations and automated filtering to prevent feedback loops, the processing of external content creates a standard surface for indirect prompt injection.
Audit Metadata