The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute the shell command uv run skills/neo-clarification/scripts/validate-requirements.py -i <path_to_saved_report>. The <path_to_saved_report> argument is a file path determined by the agent at runtime. If the agent is manipulated into generating a malicious filename, this pattern could lead to command injection or unauthorized file access.
[PROMPT_INJECTION]: The skill processes raw, potentially adversarial user input (complaints, logs, and screenshots) which serves as an entry point for indirect prompt injection.
Ingestion points: Raw user text, attached screenshots, and logs processed in the Perceive Phase (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish user-provided data from system instructions, nor is there a warning to ignore instructions embedded within the user data.
Capability inventory: The agent is authorized to write local files and execute shell commands using the uv tool (SKILL.md, scripts/validate-requirements.py).
Sanitization: The skill does not implement any validation or sanitization of the user-provided data before it is formatted into a report or passed as a CLI argument.
[EXTERNAL_DOWNLOADS]: The skill uses the uv environment manager to execute its validation script. uv is designed to automatically download and manage Python versions and dependencies from remote registries. This introduces a risk of executing unverified external code if the script's environment configuration or requirements are altered.

neo-clarification