PDS Engineering Node Status Reporting Pipeline

[jordanpadams]

Summary

Build a multi-component, AI-assisted pipeline to generate status updates for the PDS Engineering Node covering any timeframe from one month to one year. The pipeline collects normalized engineering activities from GitHub, synthesizes accomplishments, computes metrics, and tailors output for multiple audiences.

Architecture

GitHub
   |
Activity Collector (lasso-issues)
   |
activity.json
   |
+----------------------------+
|                            |
Accomplishment Skill     Metrics Skill
(pds-agent-skills)       (lasso-reports)
|                            |
+-------------+--------------+
              |
      Canonical Report
              |
      Audience Agent
      (pds-agent-skills)
              |
  HQ / MC / JPL / Public / Team

Design Principles

• Single source of truth (activity.json)
• Traceable activity — every claim linked to a GitHub artifact
• Separation of concerns — deterministic collection, deterministic metrics, AI-assisted synthesis
• Reusable outputs — canonical report feeds all audience variants

MVP Scope (B18)

1. Activity Collector — extended lasso-issues with PR, Release, and correlation support
2. Accomplishment Synthesis Skill — Claude Code skill in pds-agent-skills
3. Metrics Skill — Python utility in lasso-reports
4. Audience Tailoring Agent — Claude Code skill in pds-agent-skills

Sub-tasks

Sub-task issues will be tracked in their respective repositories and linked here.

lasso-issues

• Activity Collector: repository discovery and issue collection (existing + normalized)
• Activity Collector: PR collector
• Activity Collector: Release/Tag collector
• Activity Collector: Issue→PR→Release correlation engine
• Activity Collector: canonical activity.json export

lasso-reports

• Metrics Skill: compute quantitative metrics from activity.json

pds-agent-skills

• Accomplishment Synthesis Skill: transform activity.json into canonical Markdown report
• Audience Tailoring Agent: adapt canonical report for NASA HQ, MC, JPL, Engineering Team, and Public audiences

Out of Scope (follow-on)

• Slide Generation Agent
• lasso-issues → lasso-github rename (tracked separately)
• LLM-assisted clustering and semantic duplicate detection
• Discussions collector