skills.rest
YYG501
YYG501COMMUNITY
Explore all skills by this author →
GitHub

pmai-feedback

Analyzes current session transcripts to generate framework improvement prompts for PMAI.

4|Updated May 26, 2026
Search Tags:#transcript#root cause#codex#retrospective#session analysis#prompt generation#workflow feedback
One-click install
npx skills add https://github.com/YYG501/PM-AI-Harness --skill pmai-feedback-yyg501
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: pmai-feedback
Source: https://github.com/YYG501/PM-AI-Harness/tree/main/skills/feedback
Command: npx skills add https://github.com/YYG501/PM-AI-Harness --skill pmai-feedback-yyg501

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? When a PM notices collaboration problems during a PMAI workflow session, there is no structured way to trace the root cause and report it back to the PMAI framework repository. This Skill performs a read-only retrospective of the current session, pins a fixed evidence snapshot, compares actual behavior against confirmed design truth sources, and produces a ready-to-paste optimization prompt for the framework repo. ## Core Features & Use Cases - Precise Session Location: Locates the exact current session transcript via the host's session ID (currently verified for Codex), refusing to guess from recent files or timestamps. - Focused Retrospective with Audit Escalation: Defaults to focused analysis around the PM's feedback, upgrading to a full audit of the fixed snapshot only when explicit escalation conditions are met. - Root-Cause Attribution: Classifies each finding as consumer project issue, execution gap, skill gap, framework contract gap, host runtime limit, or insufficient evidence, with falsification criteria per finding. - Framework Handoff Prompt: Generates a self-contained Markdown prompt with session ID, transcript path, snapshot boundaries, findings, and reconciliation instructions for the PMAI generator repo. - Use Case: After a PMAI session where the AI repeatedly asked already-answered questions, the PM runs the feedback command; the Skill pinpoints the transcript, identifies which Skill rule was not enforced, and outputs a prompt the PM pastes into the PMAI framework repo for reconciliation. ## Quick Start Ask the AI to run /pmai-feedback to review this conversation and generate a framework improvement prompt for the PMAI repository.

Frequently Asked Questions about pmai-feedback

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

FAQPage Schema
How do I review a PMAI session and report workflow problems?▼

Run /pmai-feedback in the consumer repository where the session happened. It locates the exact session transcript, compares behavior against confirmed design documents, and outputs a copy-paste prompt for the PMAI framework repository.

How does session transcript root-cause analysis work in PMAI?▼

The Skill pins a fixed snapshot of the raw JSONL transcript, reads only the evidence needed to explain the PM's feedback, and attributes each finding to a consumer project issue, execution gap, skill gap, framework contract gap, host limit, or insufficient evidence.

Does pmai-feedback support Claude Code or other hosts?▼

Only Codex has a verified exact session-location path via CODEX_THREAD_ID mapping to session JSONL files. Other hosts without a verified session ID mapping cause the Skill to stop and report that retrospective is not supported.

When does the feedback skill escalate to a full session audit?▼

Full audit triggers only when the PM explicitly requests it, focused evidence conflicts with truth sources, irreversible unauthorized actions occurred, behavior spans disjointed context, or critical evidence gaps remain. Otherwise it stays in focused mode.

Can pmai-feedback modify my project or framework files?▼

No. The entire workflow is read-only. It never writes to the consumer repo, the PMAI framework repo, or the installation state; it only produces an analysis and a handoff prompt for the PM to act on.

Related Skills

Explore More Product & Management Skills→
aemgdcaemgdc
1

impeccable

Designs, critiques, and refines production-grade frontend interfaces through structured sub-commands.

Official
Advanced
stellastella
163

regression-hunt

Identify regression root causes and propose minimal regression tests.

Official
Intermediate
odylithodylith
34

odylith-compass-timeline-stream

Parse Odylith session timeline data and generate an auditable execution-stream report.

Official
Advanced
TelepotchTelepotch

edi-debug

Locate the active Claude session JSONL file and summarize its metadata.

Official
Intermediate