The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow file workflows/audit.md contains explicit instructions to execute shell commands on the host system. It invokes git rev-parse --show-toplevel to locate the project root, executes grep commands to find conditional renders within source code files, and contains platform-specific shell logic to open the generated HTML report using system utilities such as cmd.exe, open, or xdg-open.
[INDIRECT_PROMPT_INJECTION]: The audit workflow is exposed to an indirect prompt injection vector as it ingests and processes untrusted user repository files.
Ingestion points: The skill reads local project files (*.tsx, *.jsx, package.json, pyproject.toml) via shell commands and pattern matchers specified in workflows/audit.md.
Boundary markers: Absent. There are no explicit boundaries or instruction containment markers separating data from execution flow when examining client files.
Capability inventory: The agent has file system write capabilities (writing reports to the motion-audits/ directory) and local shell execution privileges (git, grep, and browser launcher scripts).
Sanitization: Absent. The code does not define sanitization routines before rendering input metadata or evaluating processed file components.

design-motion-principles