configkit/docs/roadmap/documentation.md

# User and Operator Documentation Readability Review

Reviewed revision: `08bb3a20c6a565646d98f68688d268c7bae39573`.

Status: complete. All D01–D18 revisions have been applied and reviewed together
for consistency. The catalog below preserves the proposals that guided the
work; the completion record identifies their implemented locations. The previous
consistency audit remains in Git history; its two findings were corrected in
`94d1e33`.

## Completion record

| Proposal | Implemented result |
| --- | --- |
| D01 | [README](../../README.md) leads with the loading outcome and minimal example, followed by optional tasks and platform/licensing links. |
| D02 | [Consumer guide](../consumers/README.md) starts with task routes, defaults, validation, and inspection; example background and the test inventory follow usage. |
| D03 | [Reference introduction](../config.md) supplies topic routes, YAML illustrations, path definitions, and common-versus-detailed subheadings. |
| D04 | [Sources](../config.md#sources), [results](../config.md#results-and-failures), and [resource lifetime](../config.md#completion-and-resource-lifetime) explain copying, reuse, saved inspection, and caller-owned resources. |
| D05 | [Priority](../config.md#priority-and-processing-order), [strict combination](../config.md#combining-mappings), and [defaults](../config.md#defaults-and-typed-decoding) have visible examples and retain collection exceptions. |
| D06 | [Discovery](../config.md#configuration-discovery) shows the four attempted paths and distinguishes optional absence, required absence, and selected-file failure. |
| D07 | [Imports](../config.md#explicit-imports) introduce the three selectors, option applicability, requiredness, ordering, and earlier syntax separately. |
| D08 | [Placement](../config.md#input-destinations) illustrates files and assembled YAML; [contexts](../config.md#source-resource-contexts) explain later resource access and ownership. |
| D09 | [Identified lists](../config.md#identified-lists-a-worked-example) show replace, append, and whole-record upsert before registration and policy-lifetime details. |
| D10 | [Inheritance](../config.md#sibling-inheritance) starts with parent/child settings, then cancellation and identified records before exact scheduling rules. |
| D11 | [Versions](../config.md#application-schema-versions) introduce application acceptance, a mode table, mounted headers, and the limits of observed evidence. |
| D12 | [Callbacks](../config.md#recorded-application-resolution) have task and helper tables, a derived-default example, and entry-state/lifetime guidance. |
| D13 | [Secrets](../config.md#typed-secret-references) separate lookup, hiding, and bootstrap; [bootstrap](../config.md#explicit-secrets-directory-bootstrap) explains empty existing variables and partial reports. |
| D14 | [Inspection](../config.md#layered-metadata) starts with reader needs and a default/main/overlay example; [diffs](../config.md#effective-configuration-diffs) explain snapshot flags and hidden changes. |
| D15 | [YAML](../config.md#yaml-documents), [defaults](../config.md#defaults-and-typed-decoding), [projection](../config.md#typed-projection-and-ownership), and [failures](../config.md#safe-failures) have practical examples; descriptor caching moved to the [internal overview](../internal/overview.md). |
| D16 | [Package documentation](../../doc.go) and the targeted public declarations explain purpose, paths, returned copies, and saved state while retaining precise contracts. |
| D17 | [Migration guidance](../consumers/README.md#intentional-migration-changes) leads with numbered consumer actions and explains package type identity alongside the retained old/new table. |
| D18 | [Release procedure](../release.md) defines operational terms and explains Go selection, publication statuses, and tag-object recovery without changing command blocks or gates. |

## Audience and Scope

ConfigKit is a library. Developers integrating its public API are its users;
they need different documentation from contributors maintaining the engine.
Configuration authors may know YAML without knowing Go reflection or graph
algorithms. Release maintainers are the audience for its operational runbook.
ConfigKit has no standalone application CLI or application deployment procedure
to explain.

| Material | Primary readers | Treatment in this review |
| --- | --- | --- |
| [Root README](../../README.md) | People evaluating or starting with ConfigKit | Explain outcomes and direct readers to a useful next page. |
| [Consumer guide](../consumers/README.md) | Application developers adopting or upgrading ConfigKit | Explain tasks and examples before advanced terminology. |
| [Configuration reference](../config.md) | Configuration authors and API users checking exact rules | Add context while keeping precise contracts available. |
| [Package documentation](../../doc.go) and public declaration comments | Users browsing the Go API | Explain practical effects alongside symbol names and conditions. |
| [Release procedure](../release.md) | Maintainers publishing or recovering a release | Clarify operational language while preserving the workflow. |
| [Release history](../releases) | Users considering an upgrade | Preserve historical records. Future notes can lead with user impact and required action. |
| [Attribution](../oss/attribution.md), `COPYRIGHT`, and license texts | Users checking rights and notices | Preserve formal notices and license language; no rewrite proposed. |
| Development guide, internal docs, ADRs, policies, and roadmaps | Repository contributors and decision reviewers | Outside the main rewrite scope; retain links for implementation and rationale. |

The adopted policies should not acquire local wording changes solely for this
exercise. Accepted ADR decisions and historical release records retain their
established treatment under the [documentation policy](../policy/documentation.md).

## Editorial Direction

Introduce each topic through the reader's task or the observable result.
Explain unfamiliar terms where they become useful. More words are appropriate
when they supply the background needed to understand a rule.

Follow introductory explanations with precise reference material. Retain names
readers will encounter in Go and YAML, and explain them rather than inventing
new API terminology. Use small illustrations beside the explanation and link
to maintained files for complete runnable examples. Initially, improve the
existing pages rather than introducing a new documentation tree.

These are contextual explanations, not global replacement instructions:

| Existing term | Useful explanation on first use |
| --- | --- |
| Mapping / sequence / scalar | A YAML group of named values / a list / a single value such as a string or number. |
| Precedence | Which input takes priority when different layers supply a setting. |
| Provenance / origin | Where a value came from, including application processing that changed it. |
| Explicit presence | Whether input supplied a setting, even when its value is `false`, `0`, empty, or null. |
| Detached record | A returned copy whose mutation cannot change ConfigKit's stored record. Explain borrowed filesystem ownership separately. |
| Frozen snapshot | The saved configuration state that inspection continues to describe after loading. |
| Source-local coordinates | Paths within one input document, before its destination is added. |
| Canonical projection | A consistent representation of typed values used for comparison or inspection; introduce its relevant effect before the internal term. |
| Dependency-ready order | Process a contribution after its dependencies are ready, using the documented tie-break rule when several are ready. |

## Catalog of Proposed Revisions

The identifiers below name proposals, not implementation stages. High priority
means a proposal helps readers understand a common or foundational task. Medium
priority addresses a specialized task or reference detail.

### D01 — Introduce ConfigKit through its useful outcomes

**Target:** [README introduction](../../README.md). **Priority:** High.

The feature inventory precedes the minimal example and introduces "snapshotted
mappings," "detached metadata," "audited defaulting," and "source/derivation
history" in quick succession.

Lead with what the application receives and what its author supplies. Suggested
direction: "ConfigKit loads YAML into your application's Go configuration type,
combines the inputs you choose, and lets you inspect where the final values came
from." Bring the existing minimal example forward. Follow it with short
capability summaries and links, introducing optional features when readers
select them. Keep platform and licensing navigation easy to find. Detailed
scheduling and ADR links can follow the consumer-oriented navigation.

### D02 — Make the consumer guide a route through common tasks

**Target:** [Consumer introduction and entry-point selection](../consumers/README.md#choosing-a-loading-entry-point),
the example-shape table, and Happy path. **Priority:** High.

The guide begins with a long responsibility inventory and examples named after
other applications. Its happy-path snippet immediately presents imports,
overlays, overrides, and a predeclared rules value.

Offer a short task index: load one file, supply defaults and validation, add
bundled lower-priority settings, select an overlay, and inspect a result. Point
the simplest case to the README example, then introduce additional options as
needed. Describe examples by their demonstrated task before their original
application names. Move extraction background and test-owner inventories below
usage guidance. Preserve links to maintained examples.

### D03 — Give the reference useful starting points

**Target:** [Configuration reference introduction](../config.md). **Priority:** High.

The opening says the document "owns" contracts and describes private parser
boundaries. The page mixes basic input selection with advanced lifecycle and
inspection details across a long reference.

Introduce it as the reference for accepted YAML, input combination, and loading
behavior. Add a compact topic index with routes for configuration authors and
Go API users. Introduce mappings, lists, layers, and configuration paths with
small illustrations. Separate common behavior, advanced interactions, and limits
through subheadings within the existing document. Keep detailed rules available
after their introductions; a separate general-purpose glossary is unnecessary
initially.

### D04 — Explain source selection and ownership in practical terms

**Target:** [Sources](../config.md#sources),
[single-source loading](../config.md#single-source-loading), and
[resource lifetime](../config.md#completion-and-resource-lifetime).
**Priority:** High.

"Opaque source," "consumption," "claimed once," "defensive source list," and
"publication copy" put implementation vocabulary ahead of caller actions.

Introduce sources as inputs passed to `Load`. Explain which are copied
immediately, which are read again on each load, and why a `Reader` source is
single-use. Keep the constructor table and explain who closes each resource.
Introduce the result before explaining copying mechanics. Suggested wording:
"You can change `Result.Config` after loading. Inspection methods continue to
describe the configuration as it was when the load finished." Preserve the
separate rule that borrowed filesystems remain under caller control and later
resource reads observe those filesystems.

### D05 — Teach combination rules with visible input and output

**Target:** [README philosophy](../../README.md#design-philosophy),
[layered loading](../config.md#layered-loading),
[strict composition](../config.md#strict-document-composition), and
[defaults](../config.md#defaults-and-typed-decoding). **Priority:** High.

Readers encounter "tier," "strict peers," "recursive precedence," and "typed
baseline" before seeing a combined configuration. Priority order and processing
order are also easy to confuse.

Use a small mapping example to show independent same-tier contributions and an
intentional higher-layer change. Explain same-tier conflicts. Show a priority
diagram separately from the processing timeline, explaining why defaults have
low priority while being applied during decoding. Keep the README summary brief
and link to full reference rules. Add a defaults example distinguishing omission,
explicit false/zero, an empty list, and null where the Go field permits it.
Keep collection exceptions visible; avoid a blanket "last file wins" explanation.

### D06 — Describe discovery as an ordered search

**Target:** [Configuration discovery](../config.md#configuration-discovery) and
[consumer discovery guidance](../consumers/README.md#explicit-main-file-discovery).
**Priority:** High.

"Directory-first Cartesian product," "ordered concatenation," and "optional
exhaustion" require translation into actual attempted filenames.

Show two directories and two filenames as an ordered list of candidates.
Explain that ConfigKit finishes the filenames for one directory before moving
to the next. Explain no-match behavior, required searches, and selected-file
failure. Keep the terminal selected-file error rule prominent so readers
understand why a bad selected file does not cause another search. Follow with
the existing preset table and specialized filesystem-context rules.

### D07 — Introduce imports through the three selection tasks

**Target:** [Explicit imports](../config.md#explicit-imports) and
[import defaults guidance](../consumers/README.md#import-defaults-and-selection-inspection).
**Priority:** High.

The opening combines enablement, structural controls, selector shapes,
requiredness, destinations, removed syntax, and diagnostic codes. Later passages
refer to "namespace-path lexical order" and "scoped assembly order."

Start with including a named file, searching a directory, and matching a path
pattern. Use small YAML illustrations and link to complete examples. Explain
shared options in a selector-applicability table. Separate optional missing
inputs from unreadable or invalid inputs. Explain that required path lists check
each named target, while required directory/glob searches need a selected file;
retain the empty `path: []` case. Explain observed ordering with a short dependency
example before its exact reference rules. Move legacy migration notes below the
first explanation of current syntax.

### D08 — Separate configuration placement from filesystem paths

**Target:** [Input destinations](../config.md#input-destinations),
[resource contexts](../config.md#source-resource-contexts), and
[mounted collection guidance](../consumers/README.md#input-destinations-and-mounted-collections).
**Priority:** High.

"Effective destination," "projected through that item," "nearest resolved
enclosing collection container," and "passive resource context" are difficult
without a picture of the configuration and the files on disk.

Show a tiny file tree beside a resulting YAML shape. Explain that a destination
chooses where settings appear, while an import filename resolves from its
declaring file. Walk through opening one resource: follow its value's source ID
to `SourceContext`, join `BaseDir`, and close the file the application opens.
Introduce fallback destination evidence after exact placement. Preserve
composite-origin ambiguity, caller-owned resource lifetimes, and the distinction
between saved configuration and later resource I/O.

### D09 — Explain collection identity and operations before policy lifetime

**Target:** [Collection rules](../config.md#strict-document-composition) and
[identified sequence guidance](../consumers/README.md#identified-prompt-sequences).
**Priority:** High.

Selectors, normalized field sets, type-tagged tuples, registrations, and retained
policy appear before the operation table. Readers who want to combine a list of
named records need substantial background first.

Start with a record list and an `id` field. Show before/after results for replace,
append, and upsert, including an upsert whose omitted lower fields disappear
during source assembly. Explain duplicate rejection within a strict tier. Next
introduce YAML declarations and Go registrations, including their different
final-validation guarantees. Label compound identity, exact numeric equality,
wildcard selectors, pending declarations, and record-owned policy as advanced
reference material. Preserve the existing operation and presence tables.

### D10 — Introduce inheritance with a parent and child

**Target:** [Sibling inheritance](../config.md#sibling-inheritance) and
[consumer inheritance guidance](../consumers/README.md#named-sibling-inheritance).
**Priority:** Medium.

The opening quickly moves from named siblings to identified records, root
restrictions, cancellation, and dependency scheduling.

Begin with a small `base`/`custom` mapping example and show the resulting child.
Explain "sibling" as another entry in the same enclosing mapping. Explain that
inheritance uses the settings remaining after source combination. Add
cancellation and identified-record examples next, followed by exact ordering.
Retain destination-local nested resolution, identity-field ownership,
cycle/missing-parent errors, and ordinary validation of parent records.

### D11 — Explain version checking as application-defined acceptance

**Target:** [Application versions](../config.md#application-schema-versions),
[shared checks](../config.md#shared-source-version-checks), and
[consumer version guidance](../consumers/README.md#shared-source-version-checks).
**Priority:** Medium.

"Accepted non-main declarations are consumed from their owned payload" and
"destination-prefixed application path" combine several concepts in one sentence.

Explain first that the application chooses the version field and supported
identifiers. Add a behavior table for the three modes, distinguishing missing
main headers, missing other headers, explicit null, and `NoSource`. Use a mounted
document example to separate checking a header inside a file from protecting its
assembled path. Explain that accepted non-main headers are removed before
combination. Preserve the distinctions between observing a declaration, running
a checker, and acceptance, including disclosure limits on `VersionEvidence`.

### D12 — Explain callbacks through the application's work

**Target:** [Recorded resolution](../config.md#recorded-application-resolution),
[semantic validation](../config.md#semantic-validation), and consumer task guidance.
**Priority:** High.

"Audited net resolution operations," "helper-assigned values," "covering record,"
and "post-decode entry state" obscure when to use each callback.

Introduce fixed defaults, deriving or normalizing values in `Resolve`, and
checking rules in `Validate` as separate tasks. Compare `Default`, `Normalize`,
and the two `Record*` methods, identifying which assign values. Explain that
`EntryInfo` describes callback-entry state even after the resolver changes
values. Use a small derived-value example to explain recording and dependencies.
State callback lifetime and read-only validation rules directly. Retain panic,
error-reporting, and helper-writeback behavior in the detailed reference.

### D13 — Separate secret references, hiding output, and file bootstrap

**Target:** [Typed references](../config.md#typed-secret-references),
[classification/rendering](../config.md#secret-classification-and-effective-rendering),
and [file bootstrap](../config.md#explicit-secrets-directory-bootstrap).
**Priority:** High.

"Credential-bearing fields," "sensitivity closure," "preflight/apply," and
"existing empty winner" demand familiarity with several different mechanisms.

Explain the YAML environment-variable name, the credential read during loading,
and the separate choice to hide a configuration field in inspection. Make clear
that reference names are visible unless separately classified. Explain the
optional helper as reading files into environment variables before a later load,
with a filename/name/value illustration using dummy data. Explain
`skipped_existing` with the already-set-but-empty case, then show how to read a
partial report. Preserve partial writes, lack of rollback, lookup timing,
application-owned safe messages, and visibility limits. Keep exact byte/name
limits in their reference paragraphs or tables.

### D14 — Explain inspection through the information readers need

**Target:** [Metadata](../config.md#layered-metadata),
[explanations](../config.md#historical-value-explanation),
[diffs](../config.md#effective-configuration-diffs), and their consumer examples.
**Priority:** High.

"Canonical typed projection," "decoding reconciliation," and "cross-protected
context" are demanding introductions to inspecting a setting.

Organize the introduction around the final value, its source, its earlier
definitions, and changes between two loads. Walk through a setting with a
default, supplied value, and later override. Distinguish metadata history from
value explanations. Explain `Present`, `Redacted`, and `HiddenChanged` with a
small table and hidden-change example. Keep source-versus-typed representation
details, atomic list comparisons, combined sensitivity, and unchanged-reference
credential rotation. Clarify which outputs contain values and which contain
attribution only.

### D15 — Lead YAML and Go type rules with authoring examples

**Target:** [YAML documents](../config.md#yaml-documents),
[scalars](../config.md#scalars), [schema binding](../config.md#application-schema-binding),
[decoding](../config.md#defaults-and-typed-decoding),
[projection/ownership](../config.md#typed-projection-and-ownership), and
[safe failures](../config.md#safe-failures). **Priority:** Medium.

"Private reflection descriptor graph," "paired text scalars," and "recursive
acyclic objects" precede common authoring needs. Safe failures explains
classifications before showing how to use a report.

Start with a YAML/Go field pairing and everyday examples: quote `yes` to use it
as text, write durations as strings, and use supported lowercase booleans.
Retain the scalar/defaults tables. Explain nil versus empty containers, ignored
fields, and `omitempty` through visible results. Put codec method sets and
recursive-type rules in advanced reference material. Explain diagnostic code,
configuration path, and source location with a small safe report and a link to
maintained error handling. Retain exact limits and error behavior. Move purely
internal caching or tree-storage explanations to existing internal docs where
useful, preserving all observable ownership guarantees.

### D16 — Make Go API comments understandable on their own

**Target:** [doc.go](../../doc.go), public declarations in
[lifecycle.go](../../lifecycle.go), [path.go](../../path.go),
[records.go](../../records.go), [inspection.go](../../inspection.go), and
[diff.go](../../diff.go). **Priority:** Medium.

The package comment is chiefly a feature inventory. `Resolution` introduces
"entry-state input provenance"; `Difference` introduces a "canonical changed
region"; several result comments rely on "detached" without explaining its effect.

Give major types an action-oriented first sentence followed by their precise
contracts. Illustrate configuration paths with a key/index example, and explain
returned-copy and saved-state behavior where relevant. Keep symbol-specific
errors, nil/zero behavior, ownership, and lifetime rules beside declarations.
Link to shared YAML semantics and maintained examples rather than repeating the
full configuration reference. Preserve public names and Go doc conventions.

### D17 — Turn migration prose into recognizable consumer actions

**Target:** [Migration changes](../consumers/README.md#intentional-migration-changes),
[module-path migration](../consumers/README.md#vanity-module-path-migration), and
[API cleanup](../consumers/README.md#pre-v1-public-api-cleanup).
**Priority:** Medium.

The old/new symbol table is useful, but "serialized diagnostic/provenance
expectations" and "migration evidence" assume repository history and advanced
Go package knowledge.

Lead with actions: update requirements and imports together, replace listed old
symbols, and review any stored or compared layer/code strings. Explain distinct
Go type identity through another dependency passing ConfigKit values to the
application. Retain the old/new table, historical-tag limitations, and the
difference between API and YAML changes. Keep downstream migration-status caveats
in a short background note. Do not choose a release version or claim that another
application has migrated.

### D18 — Make the release runbook easier to follow and recover from

**Target:** [Release procedure](../release.md), especially Go selection,
publication status, and recovery. **Priority:** Medium.

"Module-selected native executable," "consumer minimum," "bounded authoritative
reconciliation," and "peeled commit" require substantial release-tool knowledge.

Explain a candidate as the committed revision intended for release, the consumer
minimum as the oldest supported Go version, and the selected tool as the
ReleaseKit version recorded in `go.mod`. Use the existing Go patch example to
distinguish validation toolchain from consumer minimum. Keep command blocks
intact unless a separately verified wording change needs an accompanying
adjustment. Put exit statuses and next actions in a compact table. Explain why
the tag-object ID matters before asking the operator to compare it. Keep
recovery's exact URL, candidate, and tag identity conditions together.

Preserve pushing the candidate before publishing the tag, publication
authorization, immutable published tags, a single publication command,
and optional consumer verification. Keep CI waiting and macOS validation outside
the release gates. Plain language must not turn uncertain publication into
permission to retry, delete a published tag, or declare success.

## Priorities and Review Boundaries

The highest-impact starting points are the README and consumer entry routes
(D01–D03), combination/presence explanations (D05), and secret/inspection guidance
(D13–D14). The other proposals develop a selected task or reference topic. These
priorities are editorial recommendations, not additional release gates.

These revisions preserve existing contracts, diagnostic identifiers,
defaults, limits, security boundaries, and exceptions. Keep precise material
available after its plain-language introduction. A simpler explanation must not
imply universal last-value replacement, universal null support, transactional
loading, refreshed inspection after caller mutation, or stronger filesystem
isolation.

Implementation changes consumer-facing prose and Go declaration comments, with
the relocated descriptor-caching explanation in the existing internal overview.
Policies, accepted ADRs, release history, license notices, dependencies, and
production behavior remain unchanged. Validation includes exercising the revised
illustrations against the library, checking local links and Markdown tables,
comparing Go tokens to confirm comment-only edits, and verifying that release
command blocks remain identical to the implementation baseline `e74c26a`.

Completed validation with Go 1.26.10 and dependency downloads disabled:

- `./scripts/check.sh` passed formatting, whitespace, all tests (including local
  links and executable examples), vet, and build.
- Temporary programs outside the repository exercised the revised README,
  callbacks, YAML, discovery, imports, collection operations, placement,
  inheritance, version modes, inspection, secret bootstrap, defaults, and
  projection illustrations against the current library; all passed.
- A Go scanner comparison confirmed identical non-comment tokens in all six
  changed Go files. All 12 release command blocks are byte-for-byte unchanged.
- The six affected Markdown pages rendered successfully with the project's
  Markdown parser; all 27 tables have consistent column counts. Existing
  navigation anchors remain available.

The final consistency pass checked terminology, worked results, public comments,
cross-page links, documentation ownership, and the preserved release workflow.