Context Preserver
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
- COMMAND_EXECUTION (MEDIUM): The skill documentation instructs the user or agent to execute multiple shell scripts (e.g., resume.sh, save-context.sh, export-context.sh) and mentions starting a dev server. These scripts are not included in the provided file set, leading to unverified code execution risk.
- PERSISTENCE (MEDIUM): The skill implements background automation that triggers every 15 minutes and monitors inactivity or specific times. This represents a persistent execution pattern on the host system.
- DATA_EXPOSURE (LOW): The skill captures mental state notes, file paths, and WIP git commits. These artifacts are stored locally in the .adhd/ directory and are intended for sharing via export-context.sh, potentially exposing sensitive information or hardcoded secrets.
- INDIRECT_PROMPT_INJECTION (LOW): The skill ingests untrusted data from git logs and local context files to influence agent behavior during 'resume' operations.
- Ingestion points: .adhd/context.json, git log output.
- Boundary markers: Absent.
- Capability inventory: Executes shell scripts, writes to filesystem, performs git operations.
- Sanitization: Absent.
Audit Metadata