PROJECT ONE - A MINIMAL HARNESS

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS These two prompts are what a working harness engineer might ask an assistant when reviewing a first harness. Replace each bracketed placeholder with your own material. [START OF PROMPT] Review this Resume Review Agent lab implementation (main_loop.py, tool_registry.py, permission_gate.py) and check specifically for infinite-loop risk (is there a step limit), permission-gate correctness (can the agent escape the ./resumes folder via a crafted path like "../"), and whether logging is sufficient to reconstruct what happened after a run: [PASTE LAB CODE HERE]. Output format: a findings table (Area, Issue Found, Severity, Concrete Fix). [END OF PROMPT] [START OF PROMPT] Our Resume Review Agent lab's write_report tool currently [DESCRIBE CURRENT BEHAVIOR, e.g. "overwrites the whole reports folder" or "has no idempotency check"]: [PASTE lab write_report / tool_registry.py CODE HERE]. Propose the minimal code change needed to meet the Stage 1 acceptance criteria (stays within allowed folders, one report per resume, step limit respected per resume). Output format: a short diff or corrected function, plus a one-line note on what acceptance criterion it now satisfies. [END OF PROMPT]