Meta: owloop go end-to-end UX issues discovered during refactor test

[caoergou]

Background

During a real owloop go run on xyz-data-website (refactor 3 large API files into services), several UX/design flaws surfaced that together make the end-to-end experience confusing and wasteful.

Child issues

• Token counter resets between spec generation and autonomous loop, making total appear to decrease #72 Token counter resets between spec generation and autonomous loop, making total appear to decrease
• Spec auto-generator emits invalid specs with missing titles and self-contradictory Exclusions #73 Spec auto-generator emits invalid specs with missing titles and self-contradictory Exclusions
• Autonomous loop branch names should be meaningful instead of hash-only slugs #74 Autonomous loop branch names should be meaningful instead of hash-only slugs
• Generated HTML report is inaccurate and lacks AI summary of completed work #76 Generated HTML report is inaccurate and lacks AI summary of completed work
• Embed pre-commit and code-quality guardrails in spec template and agent prompt #77 Embed pre-commit and code-quality guardrails in spec template and agent prompt
• Gate integrity: vacuous acceptance criteria, Exclusions enforcement never worked (regression of #66), owloop metadata committed into user repos #78 Gate integrity: vacuous acceptance criteria, Exclusions enforcement never worked (Prevent acceptance-criteria commands from mutating excluded files #66 reopened), owloop metadata committed into user repos

Goal

Improve owloop go so that a user can fire off a goal, get valid specs, watch a single truthful progress/cost counter, easily identify the resulting branch, and receive an accurate, human-readable report of what was done.

Acceptance criteria

• Token counter resets between spec generation and autonomous loop, making total appear to decrease #72 closed or resolved
• Spec auto-generator emits invalid specs with missing titles and self-contradictory Exclusions #73 closed or resolved
• Autonomous loop branch names should be meaningful instead of hash-only slugs #74 closed or resolved
• Generated HTML report is inaccurate and lacks AI summary of completed work #76 closed or resolved
• Embed pre-commit and code-quality guardrails in spec template and agent prompt #77 closed or resolved
• Gate integrity: vacuous acceptance criteria, Exclusions enforcement never worked (regression of #66), owloop metadata committed into user repos #78 closed or resolved (includes re-resolving Prevent acceptance-criteria commands from mutating excluded files #66)
• A fresh end-to-end owloop go run shows a monotonically increasing token/cost counter from spec generation through implementation
• Generated specs pass a basic validation gate before the loop starts
• Branch name is readable without opening the report
• Final HTML report accurately reflects the owloop branch diff, commits, spec statuses, and contains an AI-generated summary of changes
• Commits do not include unrelated side-effect files (lockfile, logs, cache, coverage)