reflect

Records brain routing corrections with confidence scoring in a learnings markdown file.

6|1|Updated Nov 26, 2025
One-click install
npx skills add https://github.com/nathanvale/side-quest-marketplace-old --skill reflect-nathanvale
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: reflect
Source: https://github.com/nathanvale/side-quest-marketplace-old/tree/main/plugins/para-obsidian/skills/reflect
Command: npx skills add https://github.com/nathanvale/side-quest-marketplace-old --skill reflect-nathanvale

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When an AI orchestrator routes a request to the wrong skill, that mistake is usually forgotten and repeated. This Skill captures each routing correction, tracks how often the same mistake occurs, and promotes recurring patterns into permanent routing rules. ## Core Features & Use Cases - Correction Logging: Parses the original input, wrong skill, correct skill, and reason, then appends or increments entries in a Raw Correction Log table. - Semantic Pattern Matching: Treats similar inputs like "process inbox" and "triage my inbox" as the same pattern and generalizes them over time. - Threshold-Based Promotion: Automatically promotes a correction to the Active Learnings section once it occurs 3 or more times, creating a human-readable routing rule. - Use Case: The brain orchestrator routes "handle my inbox" to the wrong skill. Nathan corrects it, and this Skill logs the correction; after the third occurrence, a permanent rule routes inbox requests to the right skill. ## Quick Start Ask the brain orchestrator to record a routing correction by providing the original input, the skill it was wrongly routed to, the correct skill, and the reason for the correction.

Frequently Asked Questions about reflect

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

FAQPage Schema
How do I log a wrong skill routing decision for future learning?

Invoke the reflect skill with four arguments: the original input, the skill it was routed to, the correct skill, and the reason. It appends or increments a row in the Raw Correction Log table inside learnings.md.

How does correction-based routing learning work in a multi-skill agent?

Each correction is counted in a log, and semantically similar inputs are merged into one generalized pattern. Once a pattern reaches 3 occurrences, it is promoted to Active Learnings as a permanent routing rule.

Does the reflect skill match similar user inputs or only exact strings?

It uses semantic matching, not exact string comparison. Inputs like "process inbox" and "triage my inbox" are treated as the same pattern, and after two or more similar inputs the pattern is generalized.

When does a logged correction become an active routing rule?

A correction is promoted to the Active Learnings section only when its count reaches 3 or more. This threshold prevents one-off mistakes from becoming permanent rules.

Can the reflect skill modify files other than learnings.md?

No. The skill is restricted to a single responsibility and only reads and updates the learnings.md reference file. It processes exactly one correction per invocation and never touches other files.