context-surfing
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various CLI tools including
gh,npx, andentireto manage its environment and track session state. It also includes a local bash script,scripts/handoff-checker.sh, which is intended to be used as a session-start hook to notify the agent of pending work from previous sessions. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process handoff files (
.context-surfing/handoff-*.md) generated in previous sessions. This creates a surface for indirect prompt injection if those files were to be tampered with or contain instructions that influence the agent's behavior upon resumption. - Ingestion points:
.context-surfing/directory (processed inSKILL.mdandscripts/handoff-checker.sh). - Boundary markers: The skill provides instructions to read intent and plan artifacts verbatim to serve as a 'wave anchor' for grounding.
- Capability inventory: The agent has capabilities to execute shell commands (
entire,gh,npx) and perform file system operations. - Sanitization: No explicit sanitization or validation of the handoff file content is performed beyond the instruction to read it verbatim.
Audit Metadata