Audits and classifies media pipeline outputs against strict quality gates and eligibility properties.

Install

mkdir -p .claude/skills/lofn-qa && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/15428" && unzip -o skill.zip -d .claude/skills/lofn-qa && rm skill.zip

Installs to .claude/skills/lofn-qa

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Audit, validate, repair, and classify completed Lofn pipeline outputs. Use after music/image/video/story pipeline completion or suspicious partial runs. Do NOT use for creative generation or finalist ranking.
208 chars✓ has a “when” trigger
Advanced

Key capabilities

  • Identify output directory and modality for QA
  • Adopt a competitive-auditor stance for evaluation
  • Apply a full legacy QA procedure
  • Verify standalone music prompts for audio outputs
  • Check for original Lofn artifact granularity in full pipeline outputs
  • Save QA reports in the audited output directory

How it works

The skill follows a defined workflow to audit, validate, and classify Lofn pipeline outputs by applying a strict QA procedure and generating a report.

Inputs & outputs

You give it
Output directory and modality of a completed Lofn pipeline
You get back
QA_REPORT.md in the audited output directory

When to use lofn-qa

  • Verify music pipeline output
  • Validate story generation artifacts
  • Perform 15-point QA gate checks
  • Check pipeline output directory

About this skill

SKILL: Lofn QA — Router

This file is intentionally lean to prevent context collapse. The full tuned QA procedure is preserved byte-for-byte in references/qa_full_legacy.md and must be loaded just-in-time when an audit is actually performed.

Workflow

  1. Identify the output directory and modality.
  2. Read ../orchestration/references/adversarial_qa_stance.md to adopt the competitive-auditor stance.
  3. Read references/qa_full_legacy.md before performing any substantive QA audit.
  4. For music/audio outputs, additionally read references/eligibility_7_properties.md and references/suno_15_point_qa.md; classify ACCESSIBLE vs AMBITIOUS and run the full Suno 15-point gate.
  5. If using a structured report, copy the mold from assets/eligibility_report.template.md and add the Suno 15-Point QA Gate table.
  6. When deterministic eligibility scoring is available from explicit numeric scores, use scripts/score_eligibility.py.
  7. Apply the full QA procedure exactly as specified in references/qa_full_legacy.md; do not summarize or weaken its tuned requirements.
  8. For music/audio outputs, explicitly verify standalone Suno/Udio music prompts before any PASS verdict. This gate is non-waivable even if the parent task supplies a shorter custom checklist.
  9. For full Lofn pipeline outputs, explicitly verify original-Lofn artifact granularity: Steps 00–05 must exist as separate coordinator files (step00_aesthetics_and_genres.md ... step05_refine_medium.md), and Steps 06–10 must exist as separate per-pair files (pair_{NN}_step06_facets.md ... pair_{NN}_step10_revision_synthesis.md). Summary files or collapsed pair_{NN}_steps_06_10.md rollups alone are blocking pipeline-compliance failures.
  10. Save QA_REPORT.md in the audited output directory.
  11. If failures require rerun, follow the rerun task format from references/qa_full_legacy.md.

Non-Negotiables

  • The legacy QA text is authoritative until fully split into smaller verified references.
  • Do not remove tuned QA checks; move only after byte-for-byte preservation and validation.
  • QA stays strict. Do not loosen file, line-count, EMO-tag, header, standalone Suno prompt, contamination, child-safety, or artifact-provenance gates to preserve “creative freedom.” Missing required structure is still a blocking failure.
  • QA is not the creative brief. These checks are applied after creation to verify package readiness and diagnose failures. Creative agents should be led by Golden Seed / lineage / personality first, with QA requirements last as a hard output contract. If a file is structurally complete but generic, QA may mark REPAIR — SOUL LOSS; if a file is soulful but missing required prompt pieces, QA must still fail it structurally.
  • AMBITIOUS classification is not failure unless the route target was ACCESSIBLE.
  • Music/audio PASS requires a standalone copy-paste Suno/Udio music prompt per song. QA must fail any final music file that only contains scattered metadata such as [GENRE/TEMPO/KEY], [SONIC WORLD], [PRODUCTION NOTES], or lyrics without a dedicated ## 1. MUSIC PROMPT / [SUNO STYLE PROMPT:] section.
  • Music/audio PASS requires full Suno performance-script headers. QA must fail bare [EMO:...], prose EMO HEADER:, plain SONG FORM:, or missing [Section - EMO:... - Voice - Cue] headers, even if the agent self-check says PASS.
  • Custom parent QA instructions cannot waive the music prompt gate. If a custom checklist omits music prompts, QA must add the legacy prompt check back in and note that it did so.
  • QA reports for music must include a row or bullet named Standalone Suno/Udio music prompt present for every song/pair.
  • QA reports for full pipeline music must include a row or bullet named Original Lofn step granularity present for every pair. Use scripts/audit_lofn_pipeline_artifacts.py <run_dir> when available.
  • QA reports must check for inline validation evidence. If *.repair_attempt_*.md, *_VALIDATION_BLOCKED.md, or failed validator logs exist, QA must verify the underlying artifact was repaired and passed later validation; otherwise mark REPAIR REQUIRED.
  • Music/audio QA must run the Suno 15-Point QA Gate from references/suno_15_point_qa.md. The 7 eligibility properties alone are not enough for PASS; points 8–15 verify Suno readiness, hook survivability, Lofn specificity, anti-slop, and package readiness.

When not to use it

  • When the user wants to perform creative generation
  • When the user wants to rank finalists
  • When the user wants to summarize or weaken tuned QA requirements

Limitations

  • The legacy QA text is authoritative until fully split into smaller verified references
  • QA stays strict and does not loosen file, line-count, EMO-tag, header, standalone Suno prompt, contamination, child-safety, or artifact-provenance gates
  • Custom parent QA instructions cannot waive the music prompt gate

How it compares

This workflow provides a structured, rule-based audit for pipeline outputs, ensuring adherence to specific quality gates, unlike subjective manual review.

Compared to similar skills

lofn-qa side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
lofn-qa (this skill)01moReviewAdvanced
backtesting-frameworks172moNo flagsAdvanced
evaluating-machine-learning-models11moReviewIntermediate
evaluate-environments11moReviewIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

backtesting-frameworks

wshobson

Build robust backtesting systems for trading strategies with proper handling of look-ahead bias, survivorship bias, and transaction costs. Use when developing trading algorithms, validating strategies, or building backtesting infrastructure.

17126

evaluating-machine-learning-models

jeremylongshore

Build this skill allows AI assistant to evaluate machine learning models using a comprehensive suite of metrics. it should be used when the user requests model performance analysis, validation, or testing. AI assistant can use this skill to assess model accuracy, p... Use when appropriate context detected. Trigger with relevant phrases based on skill purpose.

13

evaluate-environments

PrimeIntellect-ai

Run and analyze evaluations for verifiers environments using prime eval. Use when asked to smoke-test environments, run benchmark sweeps, resume interrupted evaluations, compare models, inspect sample-level outputs, or produce evaluation summaries suitable for deciding next steps.

12

openjudge

agentscope-ai

>

00

generate-validation-notebook

monte-carlo-data

Generate SQL validation notebooks for dbt changes. Pass a GitHub PR URL or local dbt repo path.

00

langsmith-evaluator

dhar174

INVOKE THIS SKILL when building evaluation pipelines for LangSmith. Covers three core components: (1) Creating Evaluators - LLM-as-Judge, custom code; (2) Defining Run Functions - how to capture outputs and trajectories from your agent; (3) Running Evaluations - locally with evaluate() or auto-run v

00

Search skills

Search the agent skills registry