Enforce evidence-based verification and root-cause analysis for agentic coding tasks.
npx skills add https://github.com/uladzemer/fable-thinking --skill fable-thinking-ru
Please help me install this Agent Skill. Skill: fable-thinking-ru Source: https://github.com/uladzemer/fable-thinking/tree/main/ru Command: npx skills add https://github.com/uladzemer/fable-thinking --skill fable-thinking-ru
This skill addresses agentic failures in smaller models (Opus, Sonnet, Haiku) such as premature completion, symptom patching, and goal drift by enforcing a structured, verifiable operational discipline.
Load the fable-thinking-ru skill at the start of your session to enforce rigorous self-verification and goal-anchoring procedures.
High-intent search queries and answers about installing and using this skill.
To prevent hallucinated success in agentic coding, enforce a verification-first workflow that requires external evidence like logs or test output before marking any task complete. This structured discipline stops premature completion and forces iterative root-cause analysis.
Cascading symptom patches occur due to model drift in mid-tier models. Applying a Two-Strikes root-cause protocol stops this behavior by forcing a deep-dive debugging session when initial hypotheses fail, ensuring verifiable reasoning and goal-anchoring.
Maintain goal anchoring during multi-file refactoring by loading a strict operational discipline skill at the start of your session. This enforces external evidence-based verification and iterative procedures to prevent goal drift in mid-tier models like Sonnet or Haiku.
You need strict verification protocols for high-risk code modifications when using non-frontier models where model drift is a primary failure mode. This ensures every assumption is verified against external evidence, preventing symptom patching and premature completion.
Yes, this verification discipline is specifically designed for mid-tier models including Opus, Sonnet, and Haiku. It enforces operational discipline and verifiable reasoning protocols to mitigate the specific agentic failure modes these smaller models exhibit.
jfrometaDecompose complex software work into small, verifiable tasks with dependencies.
vobbilisReproduce nexus-db errors from logs and verify X-Scope-OrgID headers.
alffeiEnforce Red-Green-Refactor by requiring failing tests before coding.
uladzemerEnforce proof-based verification and root-cause analysis for agentic coding tasks.
