contextdb-autopilot
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell scripts (
scripts/ctx-agent.sh) and npm commands (npm run contextdb) to manage session data. These are local operational tools for the ContextDB system. - [DATA_EXPOSURE_AND_EXFILTRATION]: While the skill accesses session data stored in
.aios/context-db/, it does so for the purpose of local state persistence. It explicitly forbids the automatic injection of this data into model prompts, reducing the risk of accidental data exposure to the LLM. - [PROMPT_INJECTION]: The skill includes explicit instructions to avoid prompt injection, stating that ContextDB is no longer a prompt-injection layer and deprecating flags that previously allowed automated context loading.
- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect injection as it processes session history and external search queries. However, it mitigates this by requiring explicit retrieval steps (search followed by timeline inspection) and advising against pasting full reports into prompts.
Audit Metadata