Skip to content
Skills Library
Programming · Analysis

github-issue-dedupe

/github-issue-dedupe

Use when finding duplicate GitHub issues or checking for similar issues against a target issue. Not for closing, labeling, or modifying any duplicate issue.

Licence
Apache-2.0
Last repository change
28.09.2026

Assessment and security

90/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content25 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 28.09.2026, without references to retired models or APIs.

Licence

Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.

What it does and when to use it

Use when finding duplicate GitHub issues or checking for similar issues against a target issue. Not for closing, labeling, or modifying any duplicate issue.

Original author description

Use when finding duplicate GitHub issues or checking for similar issues against a target issue. Not for closing, labeling, or modifying any duplicate issue.

How it works

  1. Contract
  2. Inputs
  3. Procedure
  4. Failure and recovery
  5. Output
Original SKILL.md headings · 47 lines, 679 words
SKILL.md preview
---
name: github-issue-dedupe
description: 'Use when finding duplicate GitHub issues or checking for similar issues against a target issue. Not for closing, labeling, or modifying any duplicate issue.'
disable-model-invocation: true
---

# GitHub issue dedupe

## Contract

| Field | Bound contract |
|---|---|
| Trigger | Find duplicates or check for similar issues against a target GitHub issue. |
| Authority | Remote: posts at most one comment on the target issue listing duplicates; requires explicit human invocation. Preview the issue and the comment consequence before posting. No credentials, data-at-rest changes, paid actions, publishing, deployment, remote bulk mutation, or irreversible deletion beyond that comment. |
| Side effect | Optionally posts one comment on the target issue listing the duplicates, only when the human confirms. Duplicate issues themselves are never modified, closed, labeled, or deleted. |
| Done | A duplicates classification is returned: a high-confidence duplicates list, or a no-duplicates classification for negative results. |

## Inputs

- Target issue reference (owner/repo#number or URL). Required.
- GitHub repository scope to search (owner/repo). Required; defaults to the target issue's repository.
- Search terms, symptom phrases, error messages, or stack traces extracted from the target issue. Required; derived from the target issue title and body when not supplied.
- Optional: additional keyword overrides, labels to filter.

## Procedure

1. Resolve the target issue: read its title, body, labels, and comments with `gh issue view`. Extract symptom phrases, error messages, stack traces, and distinctive keywords. Done when: the target issue is resolved and its search terms are extracted.
2. Bound the search scope to the named repository. Do not search outside it unless the human explicitly supplies a broader scope. Done when: the search scope is bounded to the named repository.
3. Run a multi-strategy search over open and recently closed issues using `gh search issues` and `gh issue list`:
   - Keyword and title-similarity search from the extracted terms.
   - Error-message and stack-trace overlap search.
   - Symptom-phrase overlap search using multiple phrasings of the reported problem.
   Done when: all three search strategies have been run or recorded as non-runnable.
4. For each candidate, compare it against the target issue across these dimensions: title similarity, shared error text, overlapping symptom descriptions, and matching reproduction steps. Discard the target issue itself. Done when: every candidate is compared across all four dimensions.
5. Classify each remaining candidate as a duplicate or not. A candidate is a high-confidence duplicate only when 2 or more dimensions match the target issue. Done when: every candidate has a duplicate classification with the matching dimensions named.
6. If no candidate has 2 or more matching dimensions, return a no-duplicates classification. Do not post a comment. Done when: the no-duplicates classification is returned.
7. If one or more candidates qualify, present the duplicates list to the human with the target issue reference, each duplicate's number and title, and the matching dimensions. Done when: the duplicates list is presented to the human.
8. Only when the human confirms, post one comment on the target issue listing the high-confidence duplicates. Do not modify, close, label, or delete any duplicate issue. Done when: the comment is posted or the human declines.

## Failure and recovery

- Ambiguous or partial target issue: stop and report what is missing; do not guess the problem or invent search terms.
- Search returns no candidates: return a no-duplicates classification; this is a successful terminal, not a failure.
- No candidate has 2 or more matching dimensions: return a no-duplicates classification and no comment is posted.
- Comment post fails or is unauthorized: report the error and the unposted comment text; do not retry silently or modify any issue.
- Scope drift: if the search would widen beyond the named repository, stop and ask the human rather than expanding scope.
- Partial results: return whatever candidates were found with their matching dimensions; never pretend the done predicate holds when the search is incomplete.

## Output

A duplicates report containing the target issue reference, each high-confidence duplicate's number and title with the matching dimensions named. When the human confirms, one comment is posted on the target issue listing the duplicates. When no duplicate has 2 or more matching dimensions, a no-duplicates classification is returned and no comment is posted.
Files in this package (2)
  • SKILL.md 5 KB
  • agents/openai.yaml 1 KB

Versions

The database has no other skills named github-issue-dedupe.

Database snapshot: 13.11.2026 · package version eecfabe66a2d (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100update-dedupe
    Update the repo-local dedupe-issue-local companion skill using closed-as-duplicate signals. Use when maintainers repeatedly close issues as duplicates of the same canonical thread and that pattern should feed back into triage duplicate detection.
  • 90issues
    把项目待办与记忆放进 GitHub issues。来自审读、审计或用户的每条审查发现都可归档为独立 issue 并避免重复;列出与当前工作相关的未关闭 issue;关闭时用评论记录原问题、尝试过的方法、有效方案和最终修复。当审查发现或问题需要跟踪时(论文误导性断言、代码 bug、日后改进),在开工/续接并决定下一步时,或修复落地后需要带记录关闭 issue 时使用。需要 GitHub CLI;发布到公开仓库前会警告。用户说 "file issues"、"list issues"、"close issue"、"归档审查发现" 或 "列出待办 issue" 时也使用。
  • 100create-github-issue
    Create GitHub issues using the gh CLI. Use when the user wants to create a new issue, report a bug, request a feature, or create a task in GitHub. Trigger keywords - create issue, new issue, file bug, report bug, feature request, github issue.
  • 100gh-create-issue
    Use when asked to file, report, open, or raise a GitHub issue for a bug, feature, audit, or review finding.
  • 100filing-follow-up-issues
    Use when creating a GitHub issue, including follow-up issues found while implementing, reviewing or planning another issue, and before running gh issue create or a GitHub MCP issue-create tool.
  • 90github-issues
    Use when filing a bug, requesting a feature, creating an issue, or updating issue details. Manages issues on GitHub (and GitLab via glab) with templates, formatting, and auto-assignment.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.