learnings-summary-eval

Grade Phase 10 learnings summaries against delivery outcomes and citation standards.

1|2|Updated Apr 1, 2026
One-click install
npx skills add https://github.com/dimagi-internal/ace --skill learnings-summary-eval
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: learnings-summary-eval
Source: https://github.com/dimagi-internal/ace/tree/main/skills/learnings-summary-eval
Command: npx skills add https://github.com/dimagi-internal/ace --skill learnings-summary-eval

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the risk of biased, incomplete, or misaligned Phase 10 learnings summaries that are authored by the same LLM that ran the full opportunity lifecycle, which can lead to missed phase coverage, vague recommendations, and tone that does not reflect actual opportunity delivery outcomes.

Core Features & Use Cases

  • 5-Dimension Weighted Grading Rubric: Evaluates learnings summaries across opp-lifecycle coverage, recommendation actionability, tone calibration against actual delivery outcomes, evidence citation discipline, and forward-seeding clarity, with hard block rules for critical failures and an inflation guard to prevent inflated scores.
  • Independent Outcome Anchoring: Calibrates synthesis tone to real opp delivery evidence (UAT results, launch status, observation logs) rather than self-reported cycle grades, preventing celebratory closeout documentation for opportunities that missed key metrics or failed to launch.
  • Use Case: For ACE closeout workflows, use this Skill to validate that a learnings summary accurately reflects the full Phase 1–10 lifecycle, provides concrete, actionable improvements for the next opportunity cycle, and aligns with what the opp actually delivered.

Quick Start

Use the learnings-summary-eval skill to grade the Phase 10 learnings summary for the current opportunity closeout.

Frequently Asked Questions about learnings-summary-eval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate a closeout learnings summary for accuracy and tone misalignment?

Grade closeout learnings summaries by evaluating coverage gaps, recommendation actionability, and tone calibration against actual delivery outcomes like UAT results and observation logs. This independent evaluation prevents biased closeout documentation authored by the same lifecycle LLM.

Why does my opportunity closeout summary have vague recommendations and missed phase coverage?

Closeout summaries often miss phase coverage and contain vague recommendations when self-reported by the same LLM running the cycle. An independent grading rubric enforces strict citation and actionable recommendation standards to detect these synthesis gaps.

How do I prevent inflated tone in ACE closeout learnings summaries for missed opportunity metrics?

Prevent inflated closeout tone by anchoring tone calibration to independent outcome evidence such as launch status and UAT results, rather than self-reported cycle grades. An inflation guard blocks celebratory documentation for opportunities that failed to launch or missed key metrics.

What's the best way to validate Phase 10 learnings summaries against original PDD baselines?

Validate Phase 10 learnings summaries by cross-referencing synthesis artifacts against per-phase run state, UAT results, observation logs, and original PDD baselines. A 5-dimension weighted grading rubric enforces forward-seeding clarity and evidence citation discipline.

Does the learnings summary evaluation apply to ACE closeout workflows for Connect opportunities?

Yes, the learnings summary evaluation applies specifically to ACE closeout workflows for Connect opportunities. It independently grades the full Phase 1–10 lifecycle artifacts to ensure concrete, actionable improvements align with what the opportunity actually delivered.

When should I not use an independent grading rubric for closeout learnings summaries?

Avoid skipping independent closeout validation when a learnings summary is authored by the same LLM that ran the full opportunity lifecycle. Hard block rules for critical failures and an inflation guard are necessary to prevent missed phase coverage and tone misalignment with actual delivery outcomes.