Reassess against Repository Quality Standard v1.6.0

[trsdn]

Reassessment following trsdn/.github v1.6.0, which corrected the badge rule.

P08: partial → pass. 74 pass · 0 partial · 0 fail · 19 n/a. State stays Healthy.

This is the only result that changed, and it changed because the standard was corrected — not because anything here was relaxed. No file outside the record, its badge, and its evidence is touched.

Why it moved

v1.5.1 asked for badge images "served from the repository or a first-party source where practical". That names no condition a repository can satisfy, so P08 sat at partial with no work available that would clear it.

The v1.5.1 assessment also gave a wrong reason for preferring first-party images: it claimed img.shields.io observes every reader. Markdown rendered on github.com loads external images through GitHub's proxy, so the host sees the proxy, not the reader. I had carried that over from the W07 argument for the published site, where it does hold because a browser fetches those images directly. Badges appear only in the README, so it never applied there.

v1.6.0 turns the rule on how a value changes:

Value	Moves without a commit?	First-party image?	Result
Licence	No	No	Live third-party — Pass
Platform	No	No	Live third-party — Pass
CI status	Yes	Yes, GitHub's workflow endpoint	First-party live — Pass, preferred form
Latest release	Yes	No	Live third-party — Pass
Conformance	No	Committed, regenerated from record	Pass

The reading I checked before recording pass

The sibling session asked me not to take their pass on trust, and the criterion does turn on a genuinely new distinction, so both readings were tested against the pinned text.

The committing bullet reads "a badge image may be committed … only when a repository event regenerates it", naming licence, platform, and conformance as qualifying. That is a permission gate on committing, not an obligation to commit — it forbids committing a moving value, it does not require committing a stable one.

The competing reading, under which "everywhere else" excludes licence and platform so both must become committed images, makes "may" do the work of "must" and leaves a live third-party licence badge with no stated result at all. The v1.6.0 changelog settles it outright: "No recorded result can turn into a Fail from any of this."

So licence and platform stay live. Committing generated images for them is permitted and would gain nothing: it trades a value correct by construction for one correct only until the next regeneration.

What remains is an availability and trust dependency on img.shields.io for three of five badges. Real, worth revisiting if a first-party source appears — but under v1.6.0 not a defect, and recorded as such rather than left as a vague partial.

Changes

• .github/conformance.yml — standard_version 1.5.1 → 1.6.0, P08 → pass.
• .github/badges/conformance.svg — regenerated from the record with the v1.6.0 generator, not hand-edited. Label now reads v1.6.0 - Healthy.
• docs/self-assessment.md — header, P08 row, and a new Badge hosting (P08) section recording why the result moved, so a later reader can see it followed a rule change. The old Remaining gap section is replaced by Open gaps: none — it claimed P08 was the only criterion off pass and pointed at trsdn/.github#31 as still open, both now false.

Verification

• conformance.py --check against the v1.6.0 catalog: Healthy (na=19, pass=74), badge in sync.
• Record matches the v1.6.0 catalog exactly — 93 criteria, no missing or extra IDs (the catalog is unchanged between 1.5.1 and 1.6.0; IDs are append-only).
• scripts/check-documentation.sh green.
• No source, workflow, or site file modified, so the build and test surface is untouched.

The reusable conformance workflow checks out the standard at the tag named in the record, so CI now validates against v1.6.0.