Prioritize

node --experimental-strip-types ./scripts/issues/triage.mts single <numbers...>

Read the evidence pack JSON printed to stdout.

From-report mode (--from-report):

Read the JSON file and auto-detect report type:

• Has verdicts array → Triage decisions. Filter to issues with actionable verdicts (Valid - Needs Triage, Valid - Already Triaged, Retype - Bug, Retype - Feature Request). Skip close or request-more-info verdicts — those are already resolved.
• Has investigations array → Investigation decisions. For each issue, combine investigation findings (effort, risk, root cause, confidence) with evidence pack data.

For each qualifying issue, you need the evidence pack data. If the corresponding evidence pack is available in .work/triage/packs/, read it. Otherwise, run the single command to fetch data for the issue numbers.

Stage 1 — Gather Related Issue Counts

For each issue being evaluated, search for related issues:

gh search issues --repo gitkraken/vscode-gitlens "<significant keywords from title>" --state open --json number --limit 20

Extract 3-5 significant keywords from the issue title (skip common words like "the", "is", "not", "when", "with", "does", "after"). Count results excluding the issue itself.

Rate budget: Maximum 25 search calls per invocation. If evaluating more issues than that, prioritize issues with higher reaction counts or those from investigation reports.

Stage 2 — Evaluate Each Issue

For each issue, assess these priority signals:

Stage 3 — Recommend Resolution

Apply this decision framework:

1. Won't Fix — Issue is out of scope for GitLens OR describes behavior that is working as designed. Requires clear evidence.

2. Community Contribution — Feature request that is a good idea but too niche for the team to prioritize. Signals: low reaction count, specific/narrow use case, effort > Small.

3. Shortlist — Should be done in the near term. Signals: high reactions (11+), high severity (data loss, crash, broken workflow), many related issues (3+), OR confirmed bug from investigation with Small/Medium effort.

4. Backlog — Should be done, but lower priority or insufficient impact-to-effort ratio. Default for valid issues that don't meet shortlist criteria.

The skill does NOT auto-decide. Present the signals, recommendation, and confidence level. A human makes the final call.

Stage 4 — Draft Communications (when applicable)

Draft a comment only when the verdict is clear-cut:

• Won't Fix: Explain why (out of scope or working as designed), thank the reporter, suggest alternatives if applicable
• Community Contribution: Explain the feature is welcome as a contribution, provide relevant context about the codebase area (file paths, approach hints), link to contribution guidelines

Do NOT draft comments for shortlist or backlog — those are internal planning decisions.

Output

Produce two files in .work/triage/reports/:

1. Markdown Report

File: YYYY-MM-DD-RESOLUTION-REPORT.md

Single issue:

# Resolution Report — YYYY-MM-DD

Issues evaluated: 1

---

### [#NNNN — Title](https://github.com/gitkraken/vscode-gitlens/issues/NNNN)

- **Author**: @username (team) | @username
- **Recommendation**: Shortlist | Backlog | Won't Fix | Community Contribution
- **Confidence**: High | Medium | Low

#### Priority Signals

| Signal                | Value                  | Assessment                          |
| --------------------- | ---------------------- | ----------------------------------- |
| Reactions (thumbs-up) | N                      | Low / Moderate / High demand        |
| Related issues        | N open, N closed       | Isolated / Some impact / Recurring  |
| Severity              | [level]                | [assessment]                        |
| Estimated effort      | Small / Medium / Large | [source: investigation or inferred] |
| Age                   | N days                 | [assessment]                        |
| Last activity         | N days ago             | [assessment]                        |
| Comments              | N                      | [assessment]                        |

#### Rationale

[2-3 sentences explaining why this recommendation, citing specific signals]

#### Draft Message (if applicable)

> [Draft comment for won't-fix or community-contribution verdicts]

Batch (multiple issues):

# Resolution Report — YYYY-MM-DD

Issues evaluated: N
Source: [direct | triage decisions | investigation decisions]

## Summary

| Recommendation         | Count | Issues       |
| ---------------------- | ----- | ------------ |
| Shortlist              | N     | #NNNN, #NNNN |
| Backlog                | N     | #NNNN, #NNNN |
| Won't Fix              | N     | #NNNN        |
| Community Contribution | N     | #NNNN        |

## Shortlist (recommended for near-term work)

[Per-issue details, ordered by priority — highest reactions + severity first]

## Backlog

[Per-issue details, ordered by priority]

## Won't Fix

[Per-issue details]

## Community Contribution

[Per-issue details]

Omit recommendation sections that have no issues — the Summary table already shows the full breakdown. Each per-issue entry in a recommendation section should use the same Priority Signals table and Rationale format as the single-issue template.

2. Machine-Readable JSON

File: YYYY-MM-DD-RESOLUTIONS.json

{
  "reportId": "<uuid>",
  "generatedAt": "<ISO timestamp>",
  "source": "direct | triage | investigation",
  "sourceFile": "<path to source file if from-triage or from-investigation>",
  "resolutions": [
    {
      "issueNumber": 1234,
      "issueTitle": "...",
      "recommendation": "shortlist | backlog | wont-fix | community-contribution",
      "confidence": "High | Medium | Low",
      "prioritySignals": {
        "reactionsThumbsUp": 45,
        "relatedIssueCount": 5,
        "severity": "broken-workflow | crash | data-loss | degraded-experience | cosmetic",
        "estimatedEffort": "Small | Medium | Large | Unknown",
        "ageInDays": 840,
        "lastActivityDaysAgo": 90,
        "commentCount": 12
      },
      "rationale": "...",
      "draftMessage": "..." | null,
      "recommendedMilestone": "Shortlist | Backlog" | null,
      "recommendedLabels": []
    }
  ]
}

Generate a UUID for reportId. Write both files and confirm their paths to the user.

Chaining

This skill can be used standalone or as part of the issue workflow pipeline:

/triage recent → /investigate --from-report → /prioritize --from-report → /update-issues
/triage 5096   → /investigate 5096          → /prioritize 5096          → /update-issues
/prioritize 5096 5084 5070                    (standalone, direct mode)

Downstream: /update-issues can consume the resolutions JSON to apply recommendations to GitHub.