commit

openenv.yaml

@@ -0,0 +1,77 @@
+name: code-review-env
+version: "1.0.0"
+description: >
+  An OpenEnv reinforcement-learning environment where AI agents learn to
+  review source code for syntax errors, logic bugs, and security vulnerabilities.
+  Agents receive partial reward signals at each step, enabling curriculum learning
+  from easy syntax checks up to multi-vulnerability security audits.
+
+author: Simran Shaikh
+tags:
+  - openenv
+  - code-review
+  - security
+  - software-engineering
+  - real-world
+
+language: python
+
+tasks:
+  - id: easy_syntax
+    name: "Python Syntax Error Detection"
+    difficulty: easy
+    max_steps: 5
+    description: >
+      Identify and fix a syntax error hidden in a short Python function.
+      Graded on: correct issue_type, description accuracy, and fix validity.
+
+  - id: medium_logic
+    name: "Logic Bug: Off-by-One in Palindrome Check"
+    difficulty: medium
+    max_steps: 8
+    description: >
+      Find and fix a subtle off-by-one indexing bug in a palindrome checker.
+      Fix is verified by executing 5 deterministic test cases.
+
+  - id: hard_security
+    name: "Security Vulnerability: SQL Injection, Path Traversal & Weak Hashing"
+    difficulty: hard
+    max_steps: 10
+    description: >
+      Perform a full security audit of an authentication module containing
+      three distinct vulnerabilities. All must be identified and remediated.
+
+api:
+  reset: POST /reset
+  step:  POST /step
+  state: GET  /state
+  tasks: GET  /tasks
+
+reward:
+  min: 0.0
+  max: 1.0
+  type: continuous
+  partial_credit: true
+
+observation_space:
+  task_id: string
+  task_name: string
+  difficulty: string
+  language: string
+  code_snippet: string
+  context: string
+  step_number: integer
+  max_steps: integer
+  previous_feedback: string | null
+
+action_space:
+  identified_issues:
+    type: array
+    items:
+      line_number: integer | null
+      issue_type: "syntax_error | logic_bug | security_vulnerability | performance | style"
+      description: string
+      severity: "low | medium | high | critical"
+  suggested_fix: string | null
+  explanation: string
+  done: boolean