Documentation Revision Notes — Thickening Pass
Assessed: 2026-03-20 — Role: product leader + developer
Purpose
This document records the findings of a documentation audit conducted after the initial consolidation pass. The goal is to identify where the documentation is factually correct but insufficiently deep for a developer or content lead who needs to actually use the tool or extend it. We call this "thickening": adding the worked examples, decision context, and operational guidance that converts reference material into knowledge.
Audit findings by document
index.md (Home)
Thin: The page explains the mechanics but not the problem. A new reader doesn't
understand why a documentation team at scale needs this tool, what "large repository"
means in practice, or what a good outcome looks like. The score (report.score) is
mentioned in the API example with no explanation of what it means or how teams use it.
Workflow is invisible: The home page mentions workflows in passing ("Rules can be chained into workflows") but gives no sense of what problem they solve or when you would use them.
Revisions applied:
- Added "The problem this solves" section with a concrete scale context
- Added "Workflows" subsection in the How it works section
- Added "CI integration" one-liner with link to CLI reference
- Clarified what report.score means
rules-reference.md
Thin: All tables, no examples. A writer authoring their first rule set has no
reference point for what a valid date rule looks like, what pattern does differently
from r, or how comma-separated multi-values actually behave.
Missing gotchas: The [] operator is documented as "Contains" but rule authors
frequently use it for absence checks (which requires workflow inversion). The
multi-element XPath + equality problem is completely absent. These are the most
common sources of incorrect rules.
Workflow section redirects: "See DESIGN.md" is not sufficient for a rule author who needs to write their first workflow.
Revisions applied:
- Added "Flag examples" section with one complete rule per flag type
- Added "Operator examples" with non-obvious operators (date, r, l, s, p<N>)
- Added "Common mistakes" section: multi-element equality, comma parsing, negation patterns
- Expanded "Workflow step language" to include two worked examples inline
architecture.md
Thin: Lists the modules and their responsibilities but doesn't answer the developer's key question: where do I touch the code to make a change? The extension points (operators, parser, reporter) are implicit in DESIGN.md but not surfaced here.
Test strategy absent: The test suite has 189 tests at 95.68% coverage, organised to mirror the layer structure. This is worth documenting — it tells a new contributor where to find and add tests.
Revisions applied: - Added "Extension points" section: where to add an operator, a flag, a reporter format - Added "Test organisation" section mirroring layer structure - Added link to SRS for requirements traceability
cli-reference.md
Thin on output: No example of what a text report actually looks like. A user
integrating into CI doesn't know whether to parse the text output, use --format json,
or something else.
No CI guidance: The primary deployment target for this tool is CI pipelines, but the documentation has no example of a GitHub Actions step or equivalent.
Revisions applied: - Added "Sample text output" block showing a realistic pass/fail report - Added "CI integration" section with GitHub Actions example - Added note on exit codes and their meaning in pipeline context
What was NOT changed
- design.md: Already the most substantive document. Minor note added about Known Limitations pointing to the roadmap.
- product/assessment.md: Factual, complete, visually well-structured.
- product/roadmap.md: Well-structured; no thickening needed yet.
- srs.md: Comprehensive for the current version; will need update when v0.3 ships.
- API docs (
api/*.md): These are auto-generated from docstrings; thickening those docs means improving the source docstrings, which is a separate engineering task.