pipeline-quality-audit

pipeline-quality-audit is a skill for Claude Code from manhua-man/harness-coding-protocol. It costs 56 tokens per session (1,218 once invoked), scanned A, original, MIT.

A workflow for checking whether media or AI-generated output is actually usable, rather than trusting that a command finished successfully. It can inspect, repair, retry, fall back and retain failed attempts for diagnosis.

In plain words
What is it for?
Use it for media or AI-generation pipelines that may deliver unusable files. It helps define the required output, detect defects, normalize files, retry failed stages and preserve diagnostics.
Why use it?
A successful command can still produce a broken, empty or unsuitable file. This workflow helps recover usable output automatically and explains what happened when recovery is not possible.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the harness-coding-protocol plugin — 4 skills, 4 commands shipped together

Good fit Use it for media or AI-generation pipelines that may deliver unusable files. It helps define the required output, detect defects, normalize files, retry failed stages and preserve diagnostics.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/manhua-man/harness-coding-protocol/pipeline-quality-audit
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add manhua-man/harness-coding-protocol --skill pipeline-quality-audit
Clone the repo
git clone --depth 1 https://github.com/manhua-man/harness-coding-protocol

Made for: Claude Code.

Or install harness-coding-protocol, the plugin that ships this one along with the rest of its 4 skills, 4 commands.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for pipeline-quality-audit

README.md
[![agentmods](https://agentmods.dev/badge/skills/manhua-man/harness-coding-protocol/pipeline-quality-audit/github.svg)](https://agentmods.dev/skills/manhua-man/harness-coding-protocol/pipeline-quality-audit)
Your own site
<a href="https://agentmods.dev/skills/manhua-man/harness-coding-protocol/pipeline-quality-audit"><img src="https://agentmods.dev/badge/skills/manhua-man/harness-coding-protocol/pipeline-quality-audit/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for pipeline-quality-audit

Your own site · 80×15
<a href="https://agentmods.dev/skills/manhua-man/harness-coding-protocol/pipeline-quality-audit"><img src="https://agentmods.dev/badge/skills/manhua-man/harness-coding-protocol/pipeline-quality-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 56 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,218 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00056 $0.01218
Opus 5 $0.00028 $0.00609
Sonnet 5 $0.00011 $0.00244
Haiku 4.5 $0.00006 $0.00122

Measured 9d ago against content hash 04b919cae62d, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

pipeline-quality-audit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

ai-ide/skills/pipeline-quality-audit/SKILL.md · 100 lines

How it starts

The opening of the file, as written. The whole thing — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Pipeline Quality Audit

Judge the delivered media, not the command exit code. Prefer a recovery-first pipeline that returns a usable artifact and an observable account of repairs over one that rejects recoverable inputs and sends operational work back to the user.

Outcome contract

Define the minimum usable result from project evidence before changing the pipeline:

  • media type, dimensions/aspect ratio, duration, frame rate, and required streams;
  • content requirements such as a visible subject, intentional motion, continuity, or style anchors;
  • consumer path that must be able to open or retrieve the artifact;
  • acceptable automatic repairs, retries, fallbacks, and quality degradation.

Do not invent requirements that the caller or product does not have. A static shot, dark frame, missing audio track, or unusual aspect ratio is not a defect unless it violates the actual contract.

Recovery-first workflow

Use this order:

detect
  -> normalize or repair locally
  -> retry the failed stage with bounded changed inputs
  -> use a declared fallback or lower-cost profile
  -> quarantine the unusable attempt and retain diagnostics
  -> terminal non-delivery only when no valid artifact can be produced

For every automatic action, record the attempt, observed defect, repair/fallback chosen, resulting artifact, and final consumer probe. Do not report a repaired failure as a clean first-attempt success.

1. Detect from actual output

  • Probe the produced file with stable tools such as ffprobe, decoder reads, or image statistics.
  • Black-frame detection may use ffmpeg blackdetect: pix_th is the darkness threshold and pic_th is the proportion of dark pixels. Treat thresholds as project-calibrated signals, not universal truth.
  • Motion or duplicate-frame checks must exempt intentionally static content.
  • Validate the final delivery contract, not only an intermediate render.
  • Use numeric metrics to locate suspicious segments; use the product contract or visual review to decide whether the content is acceptable.

Read the full file on GitHub · 100 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 100 lines · 56 tokens per session scan A 04b919cae62d

Subscribe to this mod's changes

pipeline-quality-audit is a skill published in the GitHub repository manhua-man/harness-coding-protocol (5 stars, last pushed yesterday), licensed MIT. It adds 56 tokens to every session and 1,218 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

brainstorming

A brainstorming workflow for turning an idea into an agreed design before implementation. It requires exploring the project, asking clarifying questions, comparing options, documenting the design, and getting user approval.

jnMetaCode/superpowers-zh · 40 tokens

verification-before-completion

A checklist for verifying work before claiming that it is finished, fixed, built, tested, or ready to submit. It requires fresh command output and evidence for each claim.

jnMetaCode/superpowers-zh · 50 tokens

writing-skills

A guide for creating and testing reusable instructions for AI agents, called skills. It applies test-driven development, or TDD—the practice of writing tests before implementation—to instruction documents.

jnMetaCode/superpowers-zh · 23 tokens

test-driven-development

Test-driven development, or TDD, is a way to build software by writing a test that fails, adding the smallest code that makes it pass, and then cleaning up the code. These instructions require that process for features, bug fixes, refactors, and behavior changes.

jnMetaCode/superpowers-zh · 20 tokens

gentle-ai-bench

Trigger: bench, journey, journeys, driven mode, gentle-ai-bench, journey corpus, j-numbers, bench axis. Author and verify gentle-ai bench journeys; go test ./bench never proves driven execution.

Gentleman-Programming/gentle-ai · 49 tokens

go-testing

Trigger: Go tests, go test coverage, Bubbletea teatest, golden files. Apply focused Go testing patterns.

Gentleman-Programming/gentle-ai · 26 tokens