The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes internal Python scripts (setup_guidance.py) via a local runtime wrapper to perform repository health checks and install documentation pointers. These executions are performed using subprocess.run without a shell and include strict path validation to ensure execution is restricted to the skill's own directory.
[DYNAMIC_EXECUTION]: The runtime environment utilizes importlib.import_module to dynamically load a local diagnostic contract script. This is used for standardized error reporting across different environments and does not involve remote code or untrusted paths.
[INDIRECT_PROMPT_INJECTION]: The skill has the capability to read and modify existing repository files such as AGENTS.md and CLAUDE.md to insert reasoning guidance blocks. This represents a potential indirect prompt injection surface.
Ingestion points: The scripts/setup_guidance.py script reads target markdown files from the provided repository root during preflight checks and planning.
Boundary markers: The script employs specific HTML-style comments (, , and ) to delimit and manage the injected content sections.
Capability inventory: The skill is capable of executing local scripts via subprocess.run and performing atomic file writes using os.replace within the repository root.
Sanitization: The implementation performs UTF-8 validation and uses strict regular expressions to target only the specific reasoning guidance blocks, which limits the risk of the agent accidentally following instructions embedded in the processed data.

think