The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to create a central 'Project Context' file (.agents/ue5-project-context.md) that serves as the source of truth for all other agents and skills in the project. This central file is populated with data ingested from external, potentially untrusted sources.
Ingestion points: The skill reads facts from the .uproject (JSON), Config/DefaultEngine.ini, and the output of various shell commands executed for discovery.
Boundary markers: The references/context-template.md uses markdown structures to organize data but does not include explicit instructions or markers to prevent consuming agents from interpreting embedded text as instructions.
Capability inventory: The skill has the capability to write files to the repository root and execute shell commands for environment discovery.
Sanitization: There is no evidence of sanitization or validation performed on the data read from project files before it is written to the shared context file.
[DYNAMIC_EXECUTION]: The skill facilitates a workflow where one agent records 'verified' shell commands (such as build and smoke test commands) with absolute paths, and subsequent agents are instructed to execute these commands directly from the context file.
Evidence: SKILL.md mandates recording 'only commands that have actually run here, with their real absolute paths' and states the context is correct when a 'fresh agent... can run the listed build and smoke commands'.
[COMMAND_EXECUTION]: The skill relies on executing shell commands to discover the local environment, including engine versions and installation paths. While necessary for the skill's primary purpose, this involves the execution of commands derived from the local environment or provided by other skills during the initialization phase.

ue5-project-context