The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/ue_perf.py contains a toolkit function that uses importlib.import_module to dynamically load sibling modules (such as ue_stat, ue_review, and ue_audit) from a path computed relative to the script's location (../../unreal-expert/scripts). This is an expected pattern for this vendor's multi-skill environment but involves dynamic loading from non-static paths.
[COMMAND_EXECUTION]: The skill uses subprocess.run to execute various external binaries, including the project's packaged game, Unreal Engine diagnostic tools like TraceQuery, and macOS system tools like xctrace and pmset for performance and thermal monitoring.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external performance data from trace logs, CSV profiler outputs, and JSONL files. This creates a surface for indirect prompt injection if an attacker can manipulate these diagnostic inputs to influence the agent's analysis results or suggested fixes.
Ingestion points: External data is ingested through the load_frames and load_trace functions in scripts/ue_perf.py.
Boundary markers: The script does not utilize explicit boundary markers or safety instructions when processing and interpolating this external data into the agent's context.
Capability inventory: The skill performs file system writes (for backups and reports), modifies project configuration files (.ini), executes subprocesses, and interfaces with a running Unreal Editor through a remote execution module.
Sanitization: The skill correctly utilizes shlex.quote to sanitize arguments when building launch commands for the game executable.

unreal-performance