context-preservation
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external data by reading from a local state file, which serves as a potential surface for indirect prompt injection if the file is tampered with by an external process.\n
- Ingestion points: Reads state information from
.pilot-shell/state.json(SKILL.md).\n - Boundary markers: None specified; the skill does not explicitly instruct the agent to ignore instructions embedded in the state file.\n
- Capability inventory: The skill is permitted to use Bash, Read, Write, Edit, Glob, and Grep tools.\n
- Sanitization: None specified; the skill does not detail any validation or sanitization of the JSON content before restoration.\n- [COMMAND_EXECUTION]: The skill is configured to use the Bash tool, primarily for infrastructure-level tasks such as monitoring context usage and managing the serialization process.\n- [EXTERNAL_DOWNLOADS]: Mentions a GitHub repository for attribution purposes. This is a reference to a well-known service and is documented neutrally as it does not involve the execution of remote scripts.
Audit Metadata