aep-scaffold

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the Bun runtime using its official installation script (https://bun.sh/install). As Bun is a well-known JavaScript runtime service, this is a standard and expected download for the stated purpose.
  • [REMOTE_CODE_EXECUTION]: The skill uses npm and bun to fetch and install external packages including @fission-ai/openspec and better-t-stack. It also includes shell scripts for project convergence and layout auditing.
  • [COMMAND_EXECUTION]: The skill executes local shell commands to detect project stacks (e.g., checking for package.json, go.mod), manage git repositories, and verify the health of the agentic skill layout. It also executes custom scripts (scripts/audit.sh, scripts/converge.sh) to automate project setup.
  • [INDIRECT_PROMPT_INJECTION]: The skill scans project-level files to determine stack requirements and detect telemetry sources, which creates a standard ingestion surface for potentially untrusted data within the local workspace.
  • Ingestion points: Project configuration files (package.json, pyproject.toml, Cargo.toml, go.mod, turbo.json) and documentation files (AGENTS.md, CLAUDE.md).
  • Boundary markers: Not present.
  • Capability inventory: The skill performs file system modifications, executes shell commands via bash, and interacts with git and openspec CLI tools.
  • Sanitization: Not present; the tool relies on deterministic regex/grep matching for stack detection.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 11:02 AM
Security Audit — agent-trust-hub — aep-scaffold