[SKLM] Manikanta — Vibe Coding Submission

uc-0a/agents.md

@@ -1,18 +1,28 @@
-# agents.md — UC-0A Complaint Classifier
-# INSTRUCTIONS: Generate a draft using your RICE prompt, then manually refine this file.
-# Delete these comments before committing.
-
 role: >
-  [FILL IN: Who is this agent? What is its operational boundary?]
+  You are a Civic Complaint Classification Agent for the GHMC (Greater Hyderabad Municipal Corporation).
+  Your sole operational boundary is: given a single citizen complaint row, output a structured
+  classification. You do not answer questions, make recommendations, or perform any action outside
+  of classifying complaints.
 
 intent: >
-  [FILL IN: What does a correct output look like — make it verifiable]
+  A correct output is a structured record with exactly five fields:
+    - complaint_id: copied verbatim from input
+    - category: exactly one value from the allowed taxonomy (no variations, no synonyms)
+    - priority: exactly one of Urgent / Standard / Low — determined by keyword rules, not sentiment
+    - reason: one sentence citing specific words from the complaint description that justify the category and priority
+    - flag: either "NEEDS_REVIEW" or blank — set only when the category is genuinely ambiguous
+  The output is verifiable: a reviewer can check each field against the source description
+  without any additional judgment.
 
 context: >
-  [FILL IN: What information is the agent allowed to use? State exclusions explicitly.]
+  Allowed information: the complaint description field only.
+  The agent must NOT use: location names, ward numbers, reporter type, days_open, or external
+  knowledge about Hyderabad geography or infrastructure to infer category.
+  The agent must NOT hallucinate sub-categories or create category names outside the allowed list.
 
 enforcement:
-  - "[FILL IN: Specific testable rule 1 — e.g. Category must be exactly one of: Pothole, Flooding, ...]"
-  - "[FILL IN: Specific testable rule 2 — e.g. Priority must be Urgent if description contains: injury, child, school, ...]"
-  - "[FILL IN: Specific testable rule 3 — e.g. Every output row must include a reason field citing specific words from the description]"
-  - "[FILL IN: Refusal condition — e.g. If category cannot be determined from description alone, output category: Other and flag: NEEDS_REVIEW]"
+  - "Category must be exactly one of: Pothole · Flooding · Streetlight · Waste · Noise · Road Damage · Heritage Damage · Heat Hazard · Drain Blockage · Other — no spelling variations, no synonyms, no combined categories."
+  - "Priority must be set to Urgent if and only if the description contains at least one of these keywords: injury, child, school, hospital, ambulance, fire, hazard, fell, collapse, hospitalised, crater, collapsed, lives — case-insensitive match."
+  - "Every output row must include a reason field with one sentence that quotes or closely paraphrases specific words from the complaint description."
+  - "If the category cannot be determined from the description alone with confidence, set category to Other and flag to NEEDS_REVIEW — never guess a specific category."
+  - "Priority is Urgent, Standard, or Low only — never any other value. Default to Standard unless severity keywords are present (Urgent) or the complaint is minor and contains no inconvenience indicators (Low)."

uc-0a/classifier.py

@@ -1,31 +1,205 @@
 """
 UC-0A — Complaint Classifier
-Starter file. Build this using the RICE → agents.md → skills.md → CRAFT workflow.
+CRAFT-tested implementation guided by agents.md (RICE) and skills.md.
+City: Hyderabad
+
+Run:
+  python classifier.py --input ../data/city-test-files/test_hyderabad.csv --output results_hyderabad.csv
 """
+
 import argparse
 import csv
+import re
+
+# ── Taxonomy ──────────────────────────────────────────────────────────────────
+ALLOWED_CATEGORIES = [
+    "Pothole", "Flooding", "Streetlight", "Waste", "Noise",
+    "Road Damage", "Heritage Damage", "Heat Hazard", "Drain Blockage", "Other"
+]
+
+# Keywords that MUST trigger Urgent priority (case-insensitive)
+URGENT_KEYWORDS = [
+    "injury", "child", "school", "hospital", "ambulance", "fire",
+    "hazard", "fell", "collapse", "collapsed", "hospitalised",
+    "crater", "lives", "life at risk"
+]
+
+# ── Classification rules ──────────────────────────────────────────────────────
+# Each rule: (regex pattern on description, category)
+# Order matters — first match wins.
+CATEGORY_RULES = [
+    # Flooding
+    (r"\bflood(ed|ing|s)?\b|\bunderpass flood|\bwater log", "Flooding"),
+    # Drain Blockage
+    (r"\bdrain\b.*\bblock(ed)?\b|\bblock(ed)?\b.*\bdrain\b|\bstormwater drain|mosquito breeding", "Drain Blockage"),
+    # Pothole
+    (r"\bpothole(s)?\b|\bpotholes?\b", "Pothole"),
+    # Road Damage
+    (r"\broad colla?ps(e|ed)?\b|\bcra?ter\b|\broad.*damage|\bdamage.*road", "Road Damage"),
+    # Waste
+    (r"\bwaste\b|\bgarbage\b|\bwaste not cleared\b|\boverflow\b", "Waste"),
+    # Noise
+    (r"\bnoise\b|\bdrilling\b|\bidling\b|\bengines? on\b|\bsound\b", "Noise"),
+    # Heritage Damage
+    (r"\bheritage\b", "Heritage Damage"),
+    # Streetlight
+    (r"\bstreetlight(s)?\b|\blight(s)? (not working|off|broken|out)\b", "Streetlight"),
+    # Heat Hazard
+    (r"\bheat\b|\btemperature\b|\bsun(stroke)?\b", "Heat Hazard"),
+]
+
+# Low-priority signal words (when no urgent keyword and category is not critical)
+LOW_PRIORITY_SIGNALS = [r"\bslow(ed)?\b", r"\bminor\b", r"\bsmall\b"]
+
+
+def _has_urgent_keyword(text: str) -> bool:
+    text_lower = text.lower()
+    for kw in URGENT_KEYWORDS:
+        if re.search(r"\b" + re.escape(kw) + r"\b", text_lower):
+            return True
+    return False
+
+
+def _determine_category(description: str):
+    """
+    Returns (category, needs_review).
+    Matches description against CATEGORY_RULES in order.
+    Falls back to Other + NEEDS_REVIEW if no rule matches.
+    """
+    text_lower = description.lower()
+    for pattern, category in CATEGORY_RULES:
+        if re.search(pattern, text_lower):
+            return category, False
+    return "Other", True
+
+
+def _determine_priority(description: str, category: str) -> str:
+    if _has_urgent_keyword(description):
+        return "Urgent"
+    # Flooding + Drain Blockage at minimum Standard; heat/streetlight could be low
+    low_category = category in ("Streetlight", "Heat Hazard", "Noise")
+    if low_category:
+        for pattern in LOW_PRIORITY_SIGNALS:
+            if re.search(pattern, description.lower()):
+                return "Low"
+    return "Standard"
+
 
+def _build_reason(description: str, category: str, priority: str) -> str:
+    """
+    Construct a one-sentence reason citing words from the description.
+    """
+    # Find the urgent trigger word if any
+    urgent_word = None
+    for kw in URGENT_KEYWORDS:
+        if re.search(r"\b" + re.escape(kw) + r"\b", description.lower()):
+            urgent_word = kw
+            break
+
+    # Grab a short excerpt from the description (first 80 chars)
+    excerpt = description.strip()
+    if len(excerpt) > 80:
+        excerpt = excerpt[:77] + "..."
+
+    reason = f"Classified as '{category}' based on description: \"{excerpt}\""
+    if urgent_word:
+        reason += f" — priority set to Urgent due to keyword '{urgent_word}'."
+    else:
+        reason += f" — priority set to {priority}."
+    return reason
+
+
+# ── Skill: classify_complaint ─────────────────────────────────────────────────
 def classify_complaint(row: dict) -> dict:
     """
     Classify a single complaint row.
     Returns: dict with keys: complaint_id, category, priority, reason, flag
-
-    TODO: Build this using your AI tool guided by your agents.md and skills.md.
-    Your RICE enforcement rules must be reflected in this function's behaviour.
+    Implements agents.md RICE enforcement rules.
     """
-    raise NotImplementedError("Build this using your AI tool + RICE prompt")
+    complaint_id = row.get("complaint_id", "UNKNOWN")
+    description = row.get("description", "").strip()
+
+    # Handle empty description
+    if not description:
+        return {
+            "complaint_id": complaint_id,
+            "category": "Other",
+            "priority": "Low",
+            "reason": "No description provided — cannot classify.",
+            "flag": "NEEDS_REVIEW",
+        }
 
+    category, needs_review = _determine_category(description)
+    priority = _determine_priority(description, category)
+    reason = _build_reason(description, category, priority)
+    flag = "NEEDS_REVIEW" if needs_review else ""
 
+    return {
+        "complaint_id": complaint_id,
+        "category": category,
+        "priority": priority,
+        "reason": reason,
+        "flag": flag,
+    }
+
+
+# ── Skill: batch_classify ─────────────────────────────────────────────────────
 def batch_classify(input_path: str, output_path: str):
     """
     Read input CSV, classify each row, write results CSV.
-
-    TODO: Build this using your AI tool.
-    Must: flag nulls, not crash on bad rows, produce output even if some rows fail.
+    Never aborts on a bad row — logs and continues.
     """
-    raise NotImplementedError("Build this using your AI tool + RICE prompt")
+    results = []
+    errors = []
+    urgent_count = 0
+    review_count = 0
+
+    with open(input_path, newline="", encoding="utf-8") as infile:
+        reader = csv.DictReader(infile)
+        rows = list(reader)
+
+    for row in rows:
+        try:
+            result = classify_complaint(row)
+        except Exception as exc:
+            cid = row.get("complaint_id", "UNKNOWN")
+            errors.append((cid, str(exc)))
+            result = {
+                "complaint_id": cid,
+                "category": "Other",
+                "priority": "Low",
+                "reason": "Classification error — row skipped.",
+                "flag": "NEEDS_REVIEW",
+            }
+
+        results.append(result)
+        if result["priority"] == "Urgent":
+            urgent_count += 1
+        if result["flag"] == "NEEDS_REVIEW":
+            review_count += 1
+
+    fieldnames = ["complaint_id", "category", "priority", "reason", "flag"]
+    with open(output_path, "w", newline="", encoding="utf-8") as outfile:
+        writer = csv.DictWriter(outfile, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(results)
+
+    # Summary
+    print(f"\n{'='*55}")
+    print(f"  UC-0A Complaint Classifier — Results Summary")
+    print(f"{'='*55}")
+    print(f"  Total rows processed : {len(rows)}")
+    print(f"  Urgent               : {urgent_count}")
+    print(f"  NEEDS_REVIEW         : {review_count}")
+    print(f"  Classification errors: {len(errors)}")
+    if errors:
+        for cid, err in errors:
+            print(f"    ⚠ {cid}: {err}")
+    print(f"  Output written to    : {output_path}")
+    print(f"{'='*55}\n")
 
 
+# ── Entry point ───────────────────────────────────────────────────────────────
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="UC-0A Complaint Classifier")
     parser.add_argument("--input",  required=True, help="Path to test_[city].csv")

uc-0a/results_hyderabad.csv

@@ -0,0 +1,16 @@
+complaint_id,category,priority,reason,flag
+GH-202401,Flooding,Urgent,"Classified as 'Flooding' based on description: ""Underpass flooded after 1hr rain. Ambulance diverted. Lives at risk."" — priority set to Urgent due to keyword 'ambulance'.",
+GH-202402,Flooding,Standard,"Classified as 'Flooding' based on description: ""Market area flooded. Traders suffering losses. Drain completely blocked."" — priority set to Standard.",
+GH-202406,Drain Blockage,Standard,"Classified as 'Drain Blockage' based on description: ""Main stormwater drain 100% blocked with construction debris."" — priority set to Standard.",
+GH-202407,Drain Blockage,Standard,"Classified as 'Drain Blockage' based on description: ""Drain blocked and mosquito breeding. Dengue concern."" — priority set to Standard.",
+GH-202410,Pothole,Standard,"Classified as 'Pothole' based on description: ""Potholes causing vehicles to slow to 20kmph on fast road."" — priority set to Standard.",
+GH-202411,Pothole,Urgent,"Classified as 'Pothole' based on description: ""Pothole swallowed entire motorcycle wheel. Rider hospitalised."" — priority set to Urgent due to keyword 'hospitalised'.",
+GH-202412,Pothole,Urgent,"Classified as 'Pothole' based on description: ""School bus struggling to navigate 6 potholes in 200m stretch."" — priority set to Urgent due to keyword 'school'.",
+GH-202417,Waste,Standard,"Classified as 'Waste' based on description: ""Heritage zone garbage overflow. Tourist photographs showing piles of waste."" — priority set to Standard.",
+GH-202420,Noise,Standard,"Classified as 'Noise' based on description: ""Construction drilling from 5am daily near residential towers."" — priority set to Standard.",
+GH-202422,Road Damage,Urgent,"Classified as 'Road Damage' based on description: ""Road collapsed partially. Crater 1m deep near residential gate."" — priority set to Urgent due to keyword 'collapsed'.",
+GH-202424,Flooding,Standard,"Classified as 'Flooding' based on description: ""Underpass floods in light rain. Cars regularly abandoned."" — priority set to Standard.",
+GH-202428,Waste,Standard,"Classified as 'Waste' based on description: ""Post-market waste not cleared. Area unusable by Sunday morning."" — priority set to Standard.",
+GH-202432,Noise,Standard,"Classified as 'Noise' based on description: ""24hr supermarket delivery trucks idling with engines on."" — priority set to Standard.",
+GH-202448,Flooding,Standard,"Classified as 'Flooding' based on description: ""Main drain blocked — entire locality at flooding risk this week."" — priority set to Standard.",
+GH-202438,Other,Standard,"Classified as 'Other' based on description: ""Colony surrounded by fields that channel rainwater through main road."" — priority set to Standard.",NEEDS_REVIEW

uc-0a/skills.md

@@ -1,16 +1,33 @@
-# skills.md
-# INSTRUCTIONS: Generate a draft by prompting AI, then manually refine this file.
-# Delete these comments before committing.
-
 skills:
-  - name: [skill_name]
-    description: [One sentence — what does this skill do?]
-    input: [What does it receive? Type and format.]
-    output: [What does it return? Type and format.]
-    error_handling: [What does it do when input is invalid or ambiguous?]
+  - name: classify_complaint
+    description: Classifies a single civic complaint row into category, priority, reason, and flag based on RICE enforcement rules.
+    input: >
+      A dict representing one CSV row with keys:
+        complaint_id (str), description (str).
+      All other fields (ward, location, etc.) are available but must NOT be used for classification.
+    output: >
+      A dict with keys:
+        complaint_id (str) — copied verbatim from input,
+        category (str) — exactly one of: Pothole · Flooding · Streetlight · Waste · Noise · Road Damage · Heritage Damage · Heat Hazard · Drain Blockage · Other,
+        priority (str) — one of: Urgent / Standard / Low,
+        reason (str) — one sentence citing specific words from the description,
+        flag (str) — "NEEDS_REVIEW" or blank string.
+    error_handling: >
+      If description is empty or None, set category to Other, priority to Low,
+      reason to "No description provided — cannot classify", flag to "NEEDS_REVIEW".
+      Never raise an exception; always return a valid output dict.
 
-  - name: [second_skill_name]
-    description: [One sentence]
-    input: [Type and format]
-    output: [Type and format]
-    error_handling: [What does it do when input is invalid or ambiguous?]
+  - name: batch_classify
+    description: Reads an input CSV of complaints, applies classify_complaint to each row, and writes a results CSV.
+    input: >
+      input_path (str) — path to test_[city].csv,
+      output_path (str) — path to write results_[city].csv.
+      Input CSV must have columns: complaint_id, description (others may be present and are passed through).
+    output: >
+      Writes a CSV at output_path with columns:
+        complaint_id, category, priority, reason, flag.
+      Also prints a summary: total rows processed, Urgent count, NEEDS_REVIEW count, any rows that failed.
+    error_handling: >
+      If a row fails classification (unexpected exception), log the complaint_id and error,
+      write that row with category=Other, priority=Low, reason="Classification error", flag="NEEDS_REVIEW",
+      and continue processing remaining rows. Never abort the entire batch for a single bad row.

uc-0b/agents.md

@@ -1,18 +1,18 @@
 # agents.md
-# INSTRUCTIONS: Generate a draft using your RICE prompt, then manually refine this file.
-# Delete these comments before committing.
-
 role: >
-  [FILL IN: Who is this agent? What is its operational boundary?]
+  Policy Compliance Summarizer. Produces accurate, clause-complete summaries of HR policy documents.
+  Must preserve every numbered obligation exactly as written in the source.
 
 intent: >
-  [FILL IN: What does a correct output look like — make it verifiable]
+  Generate summary_hr_leave.txt containing all 10 mandatory clauses (2.3, 2.4, 2.5, 2.6, 2.7, 3.2, 3.4, 5.2, 5.3, 7.2)
+  with exact wording for multi-condition obligations. Output is verifiable against source document.
 
 context: >
-  [FILL IN: What information is the agent allowed to use? State exclusions explicitly.]
+  Use ONLY policy_hr_leave.txt as source. Do NOT add any information not present in the source.
+  Exclusions: No assumptions about "standard practice", no generic policy language.
 
 enforcement:
-  - "[FILL IN: Specific testable rule 1]"
-  - "[FILL IN: Specific testable rule 2]"
-  - "[FILL IN: Specific testable rule 3]"
-  - "[FILL IN: Refusal condition — when should the system refuse rather than guess?]"
+  - "Every numbered clause (2.3, 2.4, 2.5, 2.6, 2.7, 3.2, 3.4, 5.2, 5.3, 7.2) must be present in output"
+  - "Multi-condition obligations must preserve ALL conditions — clause 5.2 requires BOTH Department Head AND HR Director"
+  - "Never add information absent from source document — no scope bleed"
+  - "If clause cannot be summarised without meaning loss, quote verbatim and flag with [QUOTE] tag"