Cluster workflow observations into instinct files with evidence counts and confidence.
npx skills add https://github.com/softwaresalt/csv-managed --skill learn-softwaresalt
Please help me install this Agent Skill. Skill: learn Source: https://github.com/softwaresalt/csv-managed/tree/main/.github/skills/learn Command: npx skills add https://github.com/softwaresalt/csv-managed --skill learn-softwaresalt
It helps you systematically convert a high volume of workflow observations into stable, repeatable behavioral “instincts” instead of leaving patterns trapped in logs or ad-hoc notes.
Use the learn skill on the most recent observation scope to generate evidence-ranked instinct files for recurring conventions.
High-intent search queries and answers about installing and using this skill.
To turn recurring workflow observations into reusable patterns, you cluster captured observation sets by workflow phase, affected surfaces, and repeated operator corrections, producing instinct files annotated with corroborating evidence counts and confidence levels.
Observation mining for continuous learning is the process of extracting recurring practices from an observation store across sessions or release units. It systematically converts high-volume workflow logs into stable, repeatable behavioral instincts.
Confidence scoring for discovered behavioral instincts works by annotating clustered patterns with corroborating evidence counts. Low-confidence hypotheses remain explicitly provisional, distinguishing between observing a pattern and evolving it into stable instruction.
Yes, you can apply pattern discovery across different operational contexts. The clustering process targets continuous-learning workflows where patterns recur across sessions, release units, or varied operational environments by selecting the appropriate observation scope.
The best way to group observations by workflow phase and corrections is to use an evidence-based clustering approach. This method reads observation sets and groups them by phase, affected surfaces, and repeated operator corrections to safely generate evidence-ranked instincts.
Low-confidence hypotheses should remain provisional whenever evidence counts are insufficient to validate a stable pattern. Safe progression rules explicitly distinguish between observing a recurring behavior, evolving it into instruction, or evolving it into a skill.
oakossProvide a Redis-compatible in-memory data store for caching and real-time data operations.
llmh333Capture Gemini CLI tool-use sessions and convert recurring behaviors into confidence-scored instinct patterns.
CAPHTECHGenerate an observation checklist covering five baseline failure-mode checks.
hjemmesidekongenExtract atomic instincts from JSONL observations and update YAML definitions.
