skills/aymericb/skills/add-devseal/Gen Agent Trust Hub

add-devseal

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The host.sh script utilizes socat to establish a local TCP listener on the macOS host (127.0.0.1). This service is used to receive encrypted payloads from the Dev Container and return decrypted plaintext using the age utility and the macOS Secure Enclave.
  • [DYNAMIC_EXECUTION]: The unseal.sh script is designed to be invoked using shell process substitution (e.g., source <(./devseal/unseal.sh)). This pattern dynamically executes the decrypted environment variable exports directly into the current shell session, avoiding intermediate storage of sensitive data.
  • [DATA_EXPOSURE]: The system uses a 32-byte hex token for authentication between the container and the host. This token is stored in a directory on the host (~/Library/Application Support/devseal) which is bind-mounted as read-only into the container. The skill enforces restrictive file permissions (chmod 700) on the host to ensure only the local user can access the token and port information.
  • [SAFE]: The skill uses well-known security tools (age, age-plugin-se) and standard networking utilities. It explicitly includes a human-in-the-loop requirement, as decryption necessitates a Touch ID interaction on the host Mac, preventing autonomous or unauthorized access by an AI agent alone.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 09:50 PM
Security Audit — agent-trust-hub — add-devseal