chore(release): delete the committed 0.2.0 tarball, and ignore packed output - #92
chore(release): delete the committed 0.2.0 tarball, and ignore packed output#92AetherAI3 wants to merge 1 commit into
Conversation
… output `aether-agents-0.2.0.tgz` (576 KB) was committed to `main` on 2026-08-19 in #83. Six pull requests have merged since — #84, #85, #86, #87, #88 and #89 — so the file no longer contains the code its own name asserts it does. It is not in the `files` allowlist, so `npm publish` never includes it, and `release.yml` packs a fresh tarball from the tagged commit rather than reading this one. The exposure is a person: v0.2.0 is prepared but unpublished, and a repository root holding a file called `aether-agents-0.2.0.tgz` is an invitation to publish that file by hand and ship a build missing six merged changes. `*.tgz` is ignored so the next one cannot land the same way.
|
Superseded by #90, which already deletes I opened this without re-listing the open pull requests; #90 and #91 landed in the window between my last check and this branch. That is the second time today. Closing in favour of #90. The finding is recorded either way: the tarball was committed on 2026-08-19 in #83, and #84, #85, #86, #87, #88 and #89 have merged since, so the file no longer contains the code its name asserts. Nothing automated reads it — it is outside the |
aether-agents-0.2.0.tgz— 576 KB — sits at the repository root. It was committed on 2026-08-19 in #83, and six pull requests have merged since: #84, #85, #86, #87, #88 and #89. The file no longer contains the code its own name asserts it does.Nothing automated reads it. It is not in the
filesallowlist sonpm publishnever includes it, andrelease.ymlrunsnpm packagainst the tagged commit rather than reusing anything from the tree.The exposure is a person. v0.2.0 is prepared and unpublished; a repository root holding a file called
aether-agents-0.2.0.tgzis an invitation to publish that by hand and ship a build missing six merged changes. Found while establishing what "v0.2.0 is prepared" actually means for the release closure — the answer is thatmainis at 0.2.0, CI and CodeQL are green atac642363, no tag or GitHub release exists, npm still serves 0.1.0, and thenpm-productionenvironment andNPM_TOKENdo not exist, so the publish step fails closed.*.tgzis ignored so the next one cannot land the same way.