The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided free-text descriptions of 'uncertainties' and 'hypotheses' which are subsequently used to generate plans and formal 'Law' definitions. There is a risk of indirect prompt injection where malicious instructions embedded in these inputs could influence agent behavior in later phases.
Ingestion points: SKILL.md Step 2 (Enumeration of uncertainties from user input).
Boundary markers: Absent in templates and prompts to distinguish between user data and instructions.
Capability inventory: File system write operations and execution of the local validation script.
Sanitization: No sanitization is performed on the content of user-provided strings before processing.
[COMMAND_EXECUTION]: The skill executes a local Python script scripts/validate_uncertainty_plan.py to validate JSON plans. The script is bundled with the skill and performs only safe data structure and value range checks using the standard json and sys modules.

uncertainty-resolution