Record agent task outcomes and evidence into an append-only effectiveness.jsonl log.
npx skills add https://github.com/vTRKA/supervibe --skill evaluate-vtrka
Please help me install this Agent Skill. Skill: evaluate Source: https://github.com/vTRKA/supervibe/tree/main/skills/evaluate Command: npx skills add https://github.com/vTRKA/supervibe --skill evaluate-vtrka
It prevents AI coding sessions from becoming subjective by turning each agent’s claimed completion into an evidence-backed evaluation record you can review later.
Ask your AI to evaluate the most recent completed task by appending an effectiveness entry for the agent output and evidence packet.
High-intent search queries and answers about installing and using this skill.
Agent delivery outcomes are logged with evidence by recording task iterations, blockers, and confidence scores into an append-only effectiveness.jsonl file, preventing subjective claims by turning outputs into reviewable records.
Rubric-driven confidence scoring evaluates code delivery by selecting the correct confidence rubric from declared sources or artifact type mapping, then classifying the outcome as success, partial, or failed based on thresholds.
Constructing an evidence packet for agent evaluation requires gathering the task request, produced artifacts, workflow receipts, verification outputs, and context evidence to generate a classification with a gate status.
Agent effectiveness evaluation should be applied after each agent task completion or after user corrections, capturing rubric-based assessment for produced artifacts and verification evidence into the append-only log.
Verification-gated scoring requires receipt and evidence presence to classify success, partial, or failed outcomes, ensuring confidence thresholds are met before the agent output is validated.
luicabref97Generate YAML rubrics and score conversational artifacts with cited evidence.
systempromptioLog CRM and sales actions in an append-only ledger with C-### identifiers.
philo-grovesAutomate OSS proof-gating and validate evidence for supply-chain findings.
enigmaicon-engScore or compare LLM outputs with rubric-based LLM-as-judge evaluation and structured JSON results.
