The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches Godot export templates from 'github.com/godotengine/godot-builds' and the GUT testing framework from 'github.com/bitwes/Gut'. These are trusted community resources. The skill verifies the integrity of these downloads using SHA512 and SHA256 hashes.
[COMMAND_EXECUTION]: The skill frequently uses the 'subprocess' module to interface with the Godot engine. It executes the Godot binary for importing projects, running automated tests, and exporting builds. It also utilizes system tools such as 'ps', 'lsof', and 'cp' to manage concurrent process slots and clone project templates. These operations are essential for the technical director role the skill performs.
[INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect injection as it parses Godot engine logs, JUnit XML test reports, and CSV monitor data into the agent's context.
Ingestion points: Data enters the context via 'gd_stat.parse_godot_log', 'gd_stat.parse_junit', and 'gd_stat.read_monitor_csv'.
Boundary markers: The skill uses a specific 'AGENT_RESULT' envelope string to delimit the final JSON result from standard engine output.
Capability inventory: The skill can execute the Godot binary, write project files ('project.godot'), and manage local logs/captures.
Sanitization: Input is parsed into structured dictionaries rather than being treated as raw instructions, though it lacks explicit natural language filtering for embedded adversarial instructions in logs.
[DATA_EXPOSURE]: The skill interacts with Godot-specific application data paths and project settings. While it reads project configurations and logs, this access is restricted to the Godot environment and is used to provide audit reports to the agent.

godot-expert