Skip to content

0.9.0 - #40

Merged
illodev merged 1 commit into
mainfrom
chore/cut-0-9-0
Aug 8, 2026
Merged

0.9.0#40
illodev merged 1 commit into
mainfrom
chore/cut-0-9-0

Conversation

@illodev

@illodev illodev commented Aug 8, 2026

Copy link
Copy Markdown
Owner

Twelve fragments, three of them added, so semver says minor: 0.9.0.

What is in it

Added — every kind of record can be read in sequence, not only cards; raised says whether a person reported a card or an agent derived it; CI runs the checks a card declares and records what they proved.

Changed — every doctor finding names the module that produced it, and the stale-filename rule reaches memory records, managed documents and unreleased fragments rather than only cards.

Fixed — an explicit --root is checked the way a discovered one always was; the scope guard sees two agents that share an actor; the MCP byte ceiling stops overwriting a tool's own truncation flag; the hosted demo matches a search the way a real workspace does; the record views' filters survive a reload.

Security — search-local no longer ships four high advisories, and a trailing strip over an uncapped record body no longer backtracks.

One of the twelve is internal (the audit floor moving to moderate) so it is in the release record and not in the public CHANGELOG.md, which is what that visibility is for.

Why this is a pull request rather than three commits on main

Previous releases were cut straight onto main. This one goes through CI first because the tag is what publishes: release.yml runs check:release before npm publish, so a failure there is caught — but it is caught after the tag exists, and a tag is the one thing in this pipeline that cannot be taken back cleanly. Letting the Windows and macOS runners see the tree first costs one round trip.

The generated surfaces are regenerated in this commit, after the bump, rather than in a later one as in 0.8.1. That way the tag points at a tree whose version markers agree with its package.json instead of at one where they are a release behind.

The gate, run locally on this exact tree

  • check:release green: 488 + 10 tests, strict ratchet held at 474 known errors
  • pnpm audit and the consumer-tree audit both clean at moderate and above
  • the packed tarball installs and runs: Package smoke passed: @illodev/workfile@0.9.0 (install, init, Work, Docs, History, Memory, MCP, UI, read-only board)
  • doctor 0 errors 0 warnings

One thing the cut turned up

The clock rolled past raised's cutoff date while this was being prepared, so the rule went live and a baseline test that creates a card started seeing a second, unrelated finding. The card it creates now declares --raised derived, because that test is about the baseline gate and not about provenance — and a card filed from now on says how it came to be, which is the rule reaching the CLI exactly as intended.

After merging

The tag is the publish. v0.9.0 on the merge commit runs check:release again, publishes both packages to npm over OIDC, writes the MCP Registry entry, and creates the GitHub Release from the 0.9.0 section of CHANGELOG.md. Say the word and I will push it.

🤖 Generated with Claude Code

https://claude.ai/code/session_01D3LTdq3mzMAQ98rwBegGjU

Twelve fragments, three of them `added`, so semver says minor.

The release: cards can be read in sequence whatever kind they are, the record
views keep their filters, `raised` says whether a person asked for a card or an
agent derived it, and CI runs the checks a card declares and records what they
proved. Every doctor finding now names the module that produced it, and the
stale-filename rule reaches every kind rather than only cards. On the security
side, search-local no longer ships four high advisories and a trailing strip over
an uncapped body no longer backtracks.

One of the twelve is `internal` and is in the release record without being in the
public CHANGELOG, which is what that visibility is for.

`check:release` green on this tree: 488 + 10 tests, no advisories in the workspace
or in the tree a consumer resolves, and the packed tarball installs and runs as
0.9.0. The generated surfaces were regenerated after the bump rather than in a
later commit, so the tag points at a tree whose version markers agree with its
package.json.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01D3LTdq3mzMAQ98rwBegGjU
@vercel

vercel Bot commented Aug 8, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
workfile Ready Ready Preview Aug 8, 2026 12:08am
workfile-site Ready Ready Preview Aug 8, 2026 12:08am

@illodev
illodev merged commit 828339c into main Aug 8, 2026
16 checks passed
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