Where are the feedback signals stored?

Signals are written to structured Markdown files within the .pas/workspace/ directory for processing by the PAS orchestrator.

How does the skill handle a session with no issues?

If no issues are detected, the skill simply records 'No issues detected' and stops, avoiding the creation of unnecessary data or noise.

Does this skill increase token usage during my coding session?

No, it is designed to activate only during the shutdown phase, ensuring zero cost and no context interference during your productive work.

What kind of signals does the self-evaluation skill generate?

It generates four main types: Persistent Preference Updates (PPU), Output Quality Issues (OQI), Stability Gates (GATE), and Stability Anchors (STA).

Can it report issues about the PAS framework itself?

Yes, it can route framework-level feedback to GitHub issues if it detects deficiencies in PAS conventions or orchestration patterns.

PAS Self-Evaluation

Name: PAS Self-Evaluation
Author: ZoranSpirkovski

byZoranSpirkovski

0•

Productivity & Workflow

Generates structured improvement signals at agent shutdown to refine AI performance and maintain alignment with user preferences.

The self-evaluation skill is a specialized utility for the PAS (Process Automation System) framework that enables Claude to reflect on its performance after completing a task. By capturing structured feedback—such as persistent preferences, quality issues, and stability anchors—only during the shutdown phase, it ensures zero token overhead during productive work while providing actionable data for future iterations. This recursive learning loop helps maintain high output quality and ensures the AI adheres to long-term user preferences without requiring constant manual re-instruction.

Key Features

01Context-Aware Reflection: Analyzes user corrections and session outcomes to identify root causes of errors or successes.

02Structured Signal Generation: Produces standardized PPU, OQI, GATE, and STA signal types for systematic improvement.

03Framework Routing: Automatically identifies and flags framework-level issues for upstream resolution via GitHub issues.

04Saturation Prevention: Implements minimalist feedback rules to focus only on critical updates and avoid information noise.

05Zero-Cost Execution: Activates exclusively during the shutdown phase to avoid consuming tokens or context during active work.

060 GitHub stars

Use Cases

01Standardizing feedback loops across multi-agent systems within the PAS framework.

02Building a persistent knowledge base of past errors to prevent regression in complex coding tasks.

03Refining agent behavior based on repeated user style preferences or technical constraints.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zoranspirkovski/pas self-evaluation

For use in Claude.ai and ChatGPT

Key Features

01Context-Aware Reflection: Analyzes user corrections and session outcomes to identify root causes of errors or successes.

02Structured Signal Generation: Produces standardized PPU, OQI, GATE, and STA signal types for systematic improvement.

03Framework Routing: Automatically identifies and flags framework-level issues for upstream resolution via GitHub issues.

04Saturation Prevention: Implements minimalist feedback rules to focus only on critical updates and avoid information noise.

05Zero-Cost Execution: Activates exclusively during the shutdown phase to avoid consuming tokens or context during active work.

060 GitHub stars

Use Cases

01Standardizing feedback loops across multi-agent systems within the PAS framework.

02Building a persistent knowledge base of past errors to prevent regression in complex coding tasks.

03Refining agent behavior based on repeated user style preferences or technical constraints.