Skip to content

Community scaffolding, last duplicated renderer, and the remaining action bumps - #21

Merged
ivanvyd merged 2 commits into
mainfrom
feat/completeness-pass
Aug 1, 2026
Merged

Community scaffolding, last duplicated renderer, and the remaining action bumps#21
ivanvyd merged 2 commits into
mainfrom
feat/completeness-pass

Conversation

@ivanvyd

@ivanvyd ivanvyd commented Aug 1, 2026

Copy link
Copy Markdown
Owner

Supersedes #15, #16, #17, #18 and #19.

The last duplicated renderer

PackReportWriter carried a second copy of MarkdownWriter's table renderer plus its own escaper, and they had drifted: ask --format markdown states the row counts on truncation while a committed pack report did not. The same data made two different claims about how complete it was. There is now one implementation, so the drift is impossible by construction rather than guarded by a test.

Community health files — none of these existed

  • Issue forms, not markdown templates, so a report arrives with version, OS and cloud already structured. Both forms lead with a warning that questions, answers and query results can contain the reporter's own data: LakeSpeak redacts credentials but cannot know which table names are sensitive.
  • The feature form quotes the project's promise and points at GOVERNANCE.md before the first field. Scope is the most common reason to decline a request, and that is cheaper to learn before writing it up.
  • A PR template whose evidence section asks for the revert-and-watch-it-fail proof this project uses, and which ends by asking what a reviewer should be sceptical of.
  • CODEOWNERS listing security-relevant paths separately so a change to auth, redaction or the workflows is never skimmed as routine.
  • SUPPORT.md routing security reports away from public issues.

Deliberately not added: FUNDING.yml (no sponsorship) and CITATION.cff (a CLI, not something anyone cites).

Action bumps

upload-artifact and download-artifact are a matched pair — an artifact written by one major version is not readable by a mismatched other, and the release workflow writes in one job and reads in another. Dependabot split them across #19 and #15, the same split that made the CodeQL bump fail. They move together here, along with dependency-review v5, gh-release v3 and gitleaks v3. All still SHA-pinned.

147 tests green.

ivanvyd added 2 commits August 1, 2026 11:37
…renderer

PackReportWriter carried a second copy of MarkdownWriter's table renderer and its
own escaper, and the two had drifted: `ask --format markdown` stated the row counts
on truncation while a committed pack report did not, so the same data made two
different claims about how complete it was. There is now one implementation, which
makes the drift impossible rather than merely tested against.

Community health files, none of which existed:
- Issue forms rather than markdown templates, so a bug report arrives with the
  version, OS and cloud already structured. Both forms lead with a warning that
  questions, answers and query results can contain the reporter's own data —
  LakeSpeak redacts credentials but cannot know which table names are sensitive.
- The feature form quotes the project's promise and points at GOVERNANCE.md before
  the first field, because scope is the most common reason to decline a request and
  that is cheaper to learn before writing it up than after.
- A pull request template whose evidence section asks for the revert-and-watch-it-
  fail proof this project uses, and a final prompt for what a reviewer should be
  sceptical of.
- CODEOWNERS listing the security-relevant paths separately so a change to auth,
  redaction or the workflows is never skimmed as routine.
- SUPPORT.md routing security reports away from public issues, and saying plainly
  that a question about why a Genie answer is wrong is a Databricks question.

Deliberately not added: FUNDING.yml (no sponsorship) and CITATION.cff (this is a
CLI, not something anyone cites).
upload-artifact and download-artifact are a matched pair: an artifact written by
one major version is not readable by a mismatched other, and the release workflow
writes with one job and reads with another. Dependabot raised them as separate PRs
(#19 and #15), which is the same split that made the CodeQL bump fail. They move
together here.

Also folded in: dependency-review-action v5, action-gh-release v3, gitleaks-action
v3. All still pinned by commit SHA with the version in a trailing comment, so the
pins stay auditable and Dependabot can keep finding them.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant