Skip to content

fix: accuracy pass on the public surfaces before sharing (v3.4.0.1) - #69

Merged
savvides merged 1 commit into
mainfrom
fix/share-readiness
Aug 7, 2026
Merged

fix: accuracy pass on the public surfaces before sharing (v3.4.0.1)#69
savvides merged 1 commit into
mainfrom
fix/share-readiness

Conversation

@savvides

@savvides savvides commented Aug 7, 2026

Copy link
Copy Markdown
Owner

Share-readiness audit found three things wrong on public surfaces. Fixed, each with a guard so it cannot come back.

Three evidence cards claimed better evidence than the repo holds

Card Showed Actually
Needs Analysis T2–T5 all 7 references are T3
Evaluation Models T2–T5 T3–T5
Online Course Quality Frameworks T1–T5 T1–T4

The other eight were correct and every study count already matched (they sum to 108).

Overstating a tier is the one inaccuracy this project cannot ship, so the fix is not the three numbers. test/check-evidence-cards.py derives every count and tier span from evidence/references.md; smoke-test fails on disagreement, names the domain, and flags the overstating direction specifically. Two mutations pin it, one per failure axis (tier drift, count drift).

PRIVACY.md did not disclose two outbound calls

It said "No data is sent to external servers by idstack" and carved out only course-import fetching from Canvas.

  • /idstack:course-export POSTs modules, pages, assignments and discussions to your Canvas instance (skills/course-export/SKILL.md:813 onward).
  • bin/idstack-update-check runs git fetch against this repo, at most hourly.

Both now documented under Third-party services. Verified complete by enumerating every outbound host across bin/, templates/ and skills/ — the Canvas API and that git fetch are the only two.

The Windows install path could not work

README.md told Windows users to open PowerShell and run ./setup. It is an extensionless bash script and no .ps1 ships, so PowerShell cannot execute it. Now names the shell that works (WSL or Git Bash) and says why PowerShell does not.

Also

  • ROADMAP.md claimed "18 mutations" against an actual 24. Count removed rather than corrected so it cannot drift a third time.
  • TODOS.md described a "See it work" landing section that a redesign had already removed.
  • v3.4.0.0 was merged but never tagged, so release.yml never fired. Tagged retroactively at b795d02; the release is now live.

Verification

Suite Result
smoke-test.sh 354/354 (was 353)
integration-test.sh 48/48
test-setup.sh 17/17
test-doctor.sh 13/13
test-status.sh 22/22
test-manifest-merge.sh 23/23
test-version-classifier.sh 27/27
test-plugin-status.sh 9/9
test-preamble-python.sh 6/6
mutation-test.sh 24/24 guarded (was 22)

🤖 Generated with Claude Code

Three evidence cards on idstack.org claimed better evidence than the repo
holds. Needs Analysis showed T2-T5 when all 7 of its references are T3;
Evaluation Models showed T2-T5 against an actual T3-T5; Online Course
Quality Frameworks showed T1-T5 against an actual T1-T4. The other eight
cards were right and every study count already matched.

Overstating a tier is the one inaccuracy this project cannot ship, so the
fix is not the three numbers. test/check-evidence-cards.py derives every
count and tier span from evidence/references.md, smoke-test fails on any
disagreement and names the domain, and it flags the overstating direction
specifically. Two mutations pin it, one per failure axis.

PRIVACY.md said "No data is sent to external servers by idstack" and then
carved out only course-import fetching from Canvas. Two calls were
undisclosed: course-export POSTs modules, pages, assignments and
discussions to your Canvas instance, and idstack-update-check runs git
fetch against this repo hourly. Both documented. Verified complete by
enumerating every outbound host across bin/, templates/ and skills/ --
the Canvas API and that git fetch are the only two.

README's ZIP install told Windows users to open PowerShell and run
./setup. It is an extensionless bash script and no .ps1 ships, so that
path could never work. Now names WSL or Git Bash and says why.

Also: ROADMAP's "18 mutations" was 24, so the count is gone rather than
corrected; TODOS described a landing section a redesign had removed.

smoke-test 353 -> 354, mutations 22 -> 24. All ten suites green.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@savvides
savvides merged commit 67b286b into main Aug 7, 2026
8 checks passed
@savvides
savvides deleted the fix/share-readiness branch August 7, 2026 12:08
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