Bug report identification

Skill cxcscmu/SkillLearnBench/skills/b4-skill-creator-claude-haiku-4-5/github-repo-analytics/bug-report-identification

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install

npx -y skills add cxcscmu/SkillLearnBench --skill bug-report-identification

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Identify bug reports by checking if any issue label contains the substring "bug" (case-insensitive matching). Use this skill when categorizing GitHub issues as bugs vs. features, filtering defects, or generating quality/reliability metrics.

SKILL.md

3.7 KB, as published. Nobody here has run it

Bug Report Identification

Identify issues classified as bug reports based on label analysis.

Definition

A bug report is any issue where at least one label contains the substring "bug" (case-insensitive).

Examples of matching labels:

bug
type: bug
kind/bug
bug-regression
BUGFIX (case-insensitive)
p1-bug

Non-matching labels:

feature
documentation
enhancement

Algorithm

For each issue:

Extract the labels array (list of label objects with name field)
For each label, check if "bug" is a substring of the lowercased label name
If any label matches, mark the issue as a bug report
Count total bug reports and track which were closed during a specified period

Implementation (Python)

def is_bug_report(issue: Dict) -> bool:
    """
    Args:
        issue: Issue object from GitHub API with 'labels' field

    Returns:
        True if any label contains "bug" (case-insensitive)
    """
    labels = issue.get('labels', [])
    return any('bug' in label['name'].lower() for label in labels)


def count_bug_reports(issues: List[Dict]) -> int:
    """Count total issues classified as bug reports"""
    return sum(1 for issue in issues if is_bug_report(issue))


def count_closed_bugs(issues: List[Dict], closed_during: str = None) -> int:
    """
    Count bug reports that are closed.

    Args:
        issues: List of issue objects
        closed_during: Optional date range to filter (e.g., "2024-12-01..2024-12-31")
                      If None, counts all closed bugs regardless of close date

    Returns:
        Count of closed bug issues
    """
    closed_bugs = []

    for issue in issues:
        if not is_bug_report(issue):
            continue

        if issue['state'] != 'CLOSED':
            continue

        # If no date filter, include
        if closed_during is None:
            closed_bugs.append(issue)
            continue

        # If date filter provided, check closedAt
        closed_at = issue.get('closedAt')
        if closed_at:
            # Parse and compare (implementation depends on date range format)
            closed_bugs.append(issue)

    return len(closed_bugs)

Implementation (Bash with jq)

# Count total bug reports
jq '[.[] | select(.labels[].name | ascii_downcase | contains("bug"))] | length' issues.json

# Count closed bug reports
jq '[.[] | select(.labels[].name | ascii_downcase | contains("bug")) | select(.state == "CLOSED")] | length' issues.json

Edge Cases

No labels: Issue has empty labels array → not a bug report
Multiple bug-related labels: Count as one bug report (not multiple)
Label name case variation: Always normalize to lowercase for matching
Compound labels: Labels like type: bug correctly match (substring match)

Verification Steps

Manually inspect 5-10 issues marked as bugs and confirm they have relevant labels
Spot-check 2-3 non-bugs to ensure they're excluded
Count total bugs and closed bugs separately
Ensure counts don't exceed total issue count

Special Note on Date Filtering

If filtering closed bugs by date (e.g., "issues created in December, but closed when?"):

Created during period but closed anytime: No date filter on closedAt
Closed during the same period: Compare closedAt against the same date range as creation

This skill provides the counting logic; the caller should specify which date comparison to use based on the metric definition.