Lighter QA Pipeline for Documentation Issues

When a GitHub issue has documentation labels (docs, documentation, or readme), Sequant automatically uses a lighter QA pipeline optimized for content changes. The full workflow still includes spec — see #533 for the rationale on universal spec inclusion.

Prerequisites

1. Sequant installed — npx sequant --version
2. GitHub issue with a docs label — one of: docs, documentation, readme

What Changes

Documentation issues use the same three-phase workflow as other issues, but the QA phase is lighter:

Issue Type	Workflow	Phases
Standard	spec → exec → qa	3 phases
Bug fix	spec → exec → qa	3 phases (since #533)
Documentation	spec → exec → qa	3 phases (with lighter QA, see below)

The QA phase adapts when running on docs issues:

QA Behavior	Standard Issues	Docs Issues
Type safety check	Yes (sub-agent)	Skipped
Security scan	Yes (sub-agent)	Skipped
Scope/size check	Yes (sub-agent)	Yes (1 sub-agent)
Sub-agents spawned	3	1

Setup

No configuration needed. The lighter QA pipeline activates automatically when autoDetectPhases is enabled (the default).

To verify auto-detection is on:

cat .sequant/settings.json | grep autoDetectPhases
# Expected: "autoDetectPhases": true

What You Can Do

Run a documentation issue

sequant run 123

If issue #123 has a docs, documentation, or readme label, Sequant will:

1. Run spec to plan the change (since #533, spec always runs)
2. Run exec (implementation)
3. Run QA with a single sub-agent focused on scope/size and link validation (lighter than the full 3-agent QA)

Verify lighter QA is active

The orchestrator passes SEQUANT_ISSUE_TYPE=docs to the QA skill when a docs label is present. The /qa skill reads this and uses the docs-lighter pipeline (1 sub-agent instead of 3).

What to Expect

• Same spec behavior: Spec runs to plan the change, just like any other issue. This catches scope/design decisions in docs work that would otherwise slip through.
• Same exec behavior: The implementation phase runs identically.
• Lighter QA: Focuses on content accuracy, completeness, formatting, and link validity instead of type safety and security scanning.

How It Works

1. Label detection: Sequant checks issue labels exactly against docs, documentation, readme (case-insensitive equality, see exact label matching).
2. issueType propagation: If matched, SEQUANT_ISSUE_TYPE=docs environment variable is passed to post-spec phases via issueConfig.
3. QA adaptation: The QA skill reads the env var and uses a single sub-agent instead of three.

Recognized Labels

Label	Detected As
docs	Documentation
documentation	Documentation
readme	Documentation
DOCS (any case)	Documentation

Labels are matched by exact equality (case-insensitive). A label like docs-update would not trigger the lighter QA pipeline — see exact label matching for the rationale.

Troubleshooting

QA still runs 3 sub-agents on a docs issue

Symptoms: QA spawns type-safety, security, and scope agents.

Solution: This happens when running QA standalone (not via sequant run). The SEQUANT_ISSUE_TYPE env var is only set when the orchestrator manages the pipeline. Standalone /qa runs use the full pipeline by default.

Spec runs even though I want to skip it for a small docs change

This is intentional behavior since #533 — spec always runs by default because docs issues frequently contain design decisions worth a spec pass. To bypass for a specific run:

sequant run 123 --phases exec,qa

This explicit override skips auto-detection entirely.

---

Originally generated for Issue #451 on 2026-03-26; updated for #533 on 2026-04-24.