feature-pipeline

Evaluate, score, approve or reject, implement, and stage GitHub feature requests for Aries AI. Use when a GitHub issue is classified as a feature, enhancement, idea, or product request that must be co

SKILL.md

GitHub

Feature Pipeline

Use this skill for GitHub issues that represent new features, enhancements, or product requests.

Read this when needed

  • Scoring rubric: references/feature-evaluation-rubric.md

Required inputs

  • GitHub issue number
  • Repo root: /app/aries-app
  • Feedback log: data/feedback-processing-log.json

Required planning sources

Read these before deciding:

  • PRIORITIES.md
  • generated/validated/project-progress.json
  • generated/validated/canonical-roadmap-baseline.md

Required procedure

  1. Confirm the issue record is pending in the feedback log:
    • Inspect data/feedback-processing-log.json for the issue entry.
  2. Inspect the issue in GitHub:
    • gh issue view <number> --repo DeliciousHouse/aries-app --json title,body,labels,comments,url
  3. Score the request:
    • effortHours
    • impactScore
    • alignmentScore
  4. Decide approved or rejected.
    • Reject when it conflicts with the current roadmap, widens an unsupported surface, or is weakly aligned.
    • Approval here means approved for implementation and staging only, never production.
  5. If rejected:
    • update the feedback log with the reasoning
    • comment on the issue with a concise explanation
    • close the issue
  6. If approved:
    • scope the work into a tight implementation slice
    • create a branch named feature/issue-<number>-<slug>
    • implement the feature
    • run the most relevant tests
    • deploy to staging using the project's staging deploy process; if none is configured, record staging: none
    • comment on the issue with the scope summary and staging URL
    • if useful, open a draft PR for review
  7. Update data/feedback-processing-log.json so the record includes:
    • effortHours
    • impactScore
    • alignmentScore
    • approvalDecision
    • approvalReason
    • branch
    • stagingUrl
    • status
    • summaryPending

Guardrails

  • All features require Brendan's approval before production shipping.
  • Prefer rejection over stealth scope creep when the feature fights the current roadmap.
  • Keep rationale concrete. Do not close requests with generic wording.
  • Do not describe staging as production.

Output format

Return only:

  • status: success or failed
  • issue: #<number>
  • decision: approved or rejected
  • effort_hours: number
  • impact: 1-5
  • alignment: 1-5
  • branch: branch name or none
  • staging: URL or none
  • reason: short sentence
  • next_step: short operator follow-up or none