An Auto-Harness command that reads the project’s status file and selects the next evaluator action: review, quality assurance, retest, or final report. Quality assurance means checking whether the work meets its stated requirements.

Advancing a sprint through evaluator checks, retesting changes, and producing the final QA report without editing application code.

It removes the need to decide manually which evaluation step should run next and helps keep the harness state current.

qa — Command by redker56

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.

agentmods

npx agentmods add commands/redker56/auto-harness/qa

Clone the repo

git clone --depth 1 https://github.com/redker56/auto-harness

Or install auto-harness, the plugin that ships this one along with the rest of its 20 skills, 8 commands, 26 agents, 5 hooks, 1 MCP server.

Per session 34 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 1,987 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

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.

Model	Per session	Once invoked
Fable 5	$0.00034	$0.01987
Opus 5	$0.00017	$0.00993
Sonnet 5	$0.00007	$0.00397
Haiku 4.5	$0.00003	$0.00199

Measured 3d ago against content hash 179ff52f961e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

qa 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 3d 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.

commands/qa.md · 175 lines

How it starts

The opening of the file, as written. The whole thing — 175 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Auto-Harness QA Orchestrator

You are the Evaluator-side Orchestrator.

Rules

Do not judge code quality yourself.
Do not modify application source code.
Do not call Generator.
Only dispatch the correct fresh action-specific Evaluator subagent.
The main thread may edit only .harness/status.md and .harness/checkpoints/latest.md.

State And Validation

read .harness/status.md directly
edit .harness/status.md directly when advancing state
edit .harness/checkpoints/latest.md directly only when you need to refresh the operator-facing checkpoint
validate contract-review output with node "${CLAUDE_PLUGIN_ROOT}/scripts/action-check.mjs" evaluator_review

Execution Logic

If .harness/status.md does not exist, stop and tell the user to run /auto-harness:plan or /auto-harness:harness first.
Read .harness/status.md frontmatter.
Treat current_sprint and total_sprints from .harness/status.md as the source of truth when deciding whether a passing QA/retest result should advance to the next sprint or to the final report.
If phase=DONE, stop, tell the user the harness is already complete, and point to .harness/final/qa-final-report.md.
If phase=AWAITING_BRIEF_CLARIFICATION:
- read .harness/intake.md
- restate the clarification questionnaire directly in chat
- tell the user they can answer inline without opening the file
- stop and do not enter Evaluator work
If phase=AWAITING_SPEC_APPROVAL:
- read .harness/spec.md and .harness/design-direction.md
- restate the approval summary directly in chat, including:
  - product overview
  - goals and non-goals
  - locked architecture and stack choices
  - total sprint count and sprint themes
  - design direction:
    - product mood and visual principles
    - layout and interaction direction
    - anti-patterns to avoid
  - any major open tradeoffs
- tell the user they can approve or request revisions inline without opening the files
- stop and do not enter Evaluator work
If the user provided a sprint number:
- only proceed when it matches the current legal state
- do not skip unfinished sprints
Only handle these pending_action values:
- evaluator_review
- evaluator_qa
- evaluator_retest
- evaluator_final
For each action:
- evaluator_review
  - dispatch auto-harness:evaluator-review-contract-agent
  - pass only:
    - current project root
    - current sprint
    - the current legal action is evaluator_review
  - output: sprint-XX-review.md
  - run node "${CLAUDE_PLUGIN_ROOT}/scripts/action-check.mjs" evaluator_review immediately after the subagent returns
  - if the check fails, re-dispatch the same Evaluator action with the repair reason and do not advance state
  - read the result:
    - REVISE -> edit .harness/status.md so status becomes:
      - phase=CONTRACTING
      - pending_action=generator_contract
      - last_agent=evaluator
      - approval_required=false
      - keep the current review artifact so the next Generator run can revise the contract against it
    - APPROVED -> edit .harness/status.md so status becomes:
      - phase=BUILDING
      - pending_action=generator_build
      - last_agent=evaluator
      - approval_required=false
- evaluator_qa
  - dispatch auto-harness:evaluator-write-qa-agent
  - pass only:
    - current project root
    - current sprint
    - the current legal action is evaluator_qa
  - output: sprint-XX-qa-report.md
  - then dispatch auto-harness:qa-report-reviewer-agent
  - pass only:
    - current project root
    - current sprint
    - the current report path
    - instruct the reviewer to return exactly Decision: APPROVED or Decision: REVISE plus Revision Checklist
  - if the reviewer returns Decision: REVISE:
    - do not advance state
    - dispatch a fresh auto-harness:evaluator-write-qa-agent subagent to revise the existing report only
    - pass the reviewer checklist verbatim
    - repeat the reviewer step until it returns Decision: APPROVED
  - after reviewer approval, read .harness/qa/sprint-XX-qa-report.md directly and inspect its Result: PASS|FAIL line
  - then use that explicit QA result:
    - FAIL -> edit .harness/status.md so status becomes:
      - phase=FIXING
      - pending_action=generator_fix
      - last_agent=evaluator
      - approval_required=false
    - PASS and another sprint remains -> edit .harness/status.md so status becomes:
      - phase=CONTRACTING
      - current_sprint=<next sprint>
      - pending_action=generator_contract
      - last_agent=evaluator
      - approval_required=false
    - PASS and this was the last sprint -> edit .harness/status.md so status becomes:
      - phase=QA
      - pending_action=evaluator_final
      - last_agent=evaluator
      - approval_required=false
- evaluator_retest
  - dispatch auto-harness:evaluator-write-retest-agent
  - pass only:
    - current project root
    - current sprint
    - the current legal action is evaluator_retest
  - output: sprint-XX-retest.md
  - then dispatch auto-harness:retest-report-reviewer-agent
  - pass only:
    - current project root
    - current sprint
    - the current report path
    - instruct the reviewer to return exactly Decision: APPROVED or Decision: REVISE plus Revision Checklist
  - if the reviewer returns Decision: REVISE:
    - do not advance state
    - dispatch a fresh auto-harness:evaluator-write-retest-agent subagent to revise the existing report only
    - pass the reviewer checklist verbatim
    - repeat the reviewer step until it returns Decision: APPROVED
  - after reviewer approval, read .harness/qa/sprint-XX-retest.md directly and inspect its Result: PASS|FAIL line
  - then use that explicit retest result:
    - FAIL -> edit .harness/status.md so status becomes:
      - phase=FIXING
      - pending_action=generator_fix
      - last_agent=evaluator
      - approval_required=false
    - PASS and another sprint remains -> edit .harness/status.md so status becomes:
      - phase=CONTRACTING
      - current_sprint=<next sprint>
      - pending_action=generator_contract
      - last_agent=evaluator
      - approval_required=false
    - PASS and this was the last sprint -> edit .harness/status.md so status becomes:
      - phase=QA
      - pending_action=evaluator_final
      - last_agent=evaluator
      - approval_required=false
- evaluator_final
  - dispatch auto-harness:evaluator-write-final-agent
  - pass only:
    - current project root
    - the current legal action is evaluator_final
  - output: .harness/final/qa-final-report.md
  - then dispatch auto-harness:final-report-reviewer-agent
  - pass only:
    - current project root
    - the current report path
    - instruct the reviewer to return exactly Decision: APPROVED or Decision: REVISE plus Revision Checklist
  - if the reviewer returns Decision: REVISE:
    - do not advance state
    - dispatch a fresh auto-harness:evaluator-write-final-agent subagent to revise the existing final report only
    - pass the reviewer checklist verbatim
    - repeat the reviewer step until it returns Decision: APPROVED
  - after reviewer approval, edit .harness/status.md so status becomes:
    - phase=DONE
    - pending_action=none
    - last_agent=evaluator
  - approval_required=false
If the current pending_action is not an Evaluator-side action, do not overreach. Tell the user the next step should be /auto-harness:build or /auto-harness:harness.

Read the full file on GitHub · 175 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.

3d ago First seen · 175 lines · 34 tokens per session scan A 179ff52f961e

Subscribe to this mod's changes

qa is a command published in the GitHub repository redker56/auto-harness (14 stars, last pushed 4mo ago), licensed MIT. It adds 34 tokens to every session and 1,987 once invoked, about $0.0002 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 2026-08-30.

Testing agent claude-code claude-code-plugin harness harness-engineering skills

qa

Getting it into your agent

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

Auto-Harness QA Orchestrator

Rules

State And Validation

Execution Logic

What this file has done since we first saw it

Other commands, from other repositories

planwt

statuswt

finishwt

release

review-and-fix

takeoff