Proposal: visible skillshare/ project directory (#256)

[salmonumbrella]

Type

• Bug fix
• Small improvement (docs, typo, minor refactor)
• Feature proposal (proposals/ only — see CONTRIBUTING.md)

Linked Issue

#256

Summary

This PR adds a feature proposal for recognizing a visible skillshare/ directory as an alternative project marker, alongside the existing .skillshare/.

Project mode is identified solely by .skillshare/config.yaml, and trash/, backups/, logs/ and audit-rules.yaml anchor to that same hidden directory. sources (#153 / #162) made project content relocatable for repositories that want skills reviewed alongside contributor documentation, but the marker and config.yaml stayed hidden — so those repositories end up with a visible skills directory plus a hidden directory whose only job is to hold the config.

The proposal resolves the marker in a fixed order, .skillshare/config.yaml then skillshare/config.yaml. First hit wins, so .skillshare/ stays the default, takes precedence when both exist, and no migration is needed.

Notes

This PR only adds a proposal under proposals/ and does not include implementation changes, per CONTRIBUTING.md.

I did build the change first so the proposal describes real behaviour rather than a guess, and two findings from that are written into the proposal because they are easy to miss:

• The visible name collides with the global config directory. BaseDir() is <config-home>/skillshare, so a directory named skillshare already means "global config". The two places that infer scope from a config path by directory name (oplog.LogDir / ensureProjectLogGitignore, and isProjectLogConfig) have to exclude the global location by full path instead. Missing this misroutes global operation logs, and TestLogDir_Global passes a fixture path that hides the problem unless it declares a config home.
• The new package has to be a leaf. internal/audit cannot import internal/config, since config → install → audit would cycle.

The working branch is salmonumbrella:feat/visible-project-dir if it is useful as a reference — 14 non-test files plus a new internal/projectdir package, with tests for hidden-only, visible-only, both-present and neither-present, plus state-directory and init -p --visible coverage. gofmt, go vet ./... and go build ./... are clean, and the unit packages show no failures beyond the two that already fail on main outside the devcontainer (TestCommitSourceFiles_CommitFailureIsReturned, TestGitRootMismatch). Happy to leave it, drop it, or split it however you prefer — the proposal stands on its own.

Checklist

• I've read CONTRIBUTING.md
• Tests included and passing (make check) — for code changes
• No unrelated changes in the diff
• Scope is focused — one concern per PR

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a79ac35de0

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[salmonumbrella]

Converting this to a direct implementation PR: #258.

Everything in this proposal is folded into that PR body — the problem, the resolution order, the two name collisions, and the back-compat guarantees — so this document would only duplicate it. #258 also carries the working code, tests, docs and schema updates.

Issue #256 stays open as the motivation anchor.

Sorry for the shuffle. If you would rather review a proposal document first, per CONTRIBUTING, say the word and I will reopen this one and park the implementation.