Skip to content

v0.9.0

Choose a tag to compare

@github-actions github-actions released this 17 Aug 12:38
· 321 commits to main since this release

Capshelf 0.9 makes every command that describes an mcp or subagents item
say which runtimes it covers. Either kind can carry a Claude source, a Codex
source, or both, and 0.8 printed one output path for all three cases, so a
Claude-only MCP server printed .mcp.json and said nothing about Codex. This
release also finishes moving add onto the pinned commit: fragments were the
last kind whose install target set came from the data repo's working tree while
apply took it from the lock. The lock format is unchanged — 0.8 and 0.9 read
each other's locks, and there is no migration; run capshelf update skills/capshelf once per project to pick up the refreshed bundled skill. Full
details in docs/whats-new-0.9.md.

Breaking changes

  • add installs every fragment target the pinned commit holds.

    It used to pick targets from the data repo's working tree while apply picked
    them from the lock, and the cleanliness gate cannot see a deletion that
    removes a canonical path. An mcp item with codex.toml deleted in the
    worktree was therefore accepted, pinned to a commit that still contained the
    file, and installed with only .mcp.json written.

    A project that depended on that behavior will now get .codex/config.toml
    written where 0.8 wrote only .mcp.json. Commit the deletion in the data repo
    if the target is genuinely retired.

  • A canonical path committed as a directory or an empty tree is refused.

    Both used to install, save a lock, and fail in a later apply with a parser
    error, leaving a broken lock already on disk. They are refused at pin time
    now.

Enhancements

  • Report runtime target coverage on add, show, status, and share: every candidate target marked present or absent, with the output path it feeds and the canonical source path a missing one reads, read at the pinned commit rather than the worktree.
  • Add targetCoverage to add, show, share, and every status --json row, carrying present: true | false | null plus coverageState when unknown; sources, dst, and the subagent targets array keep their existing shapes.
  • Cover target coverage in the bundled capshelf skill, which guided agents through the mcp and subagent workflows without mentioning the report those commands print.

Bug fixes

  • Fire the Codex project-trust warning only when an mcp item's locked commit has a Codex source; it used to fire for every mcp item, so a Claude-only item warned about the harness it does not reach and said nothing about the one it does not cover. codex-config items keep the unconditional warning, and unknown coverage still warns — the gate removes a warning only on evidence.
  • Prove a pinned fragment source is readable before consent, so an unreadable object no longer plans as an empty contribution and locks as a covered target.
  • Stop counting a committed symlink as a covered target; Git stores one as a blob with mode 120000, so filtering on object type reported it as a present source for an item capshelf refuses to pin outright.
  • Report every output rm reconciled instead of the first, so removing a two-source mcp item names .codex/config.toml as well as .mcp.json.
  • Fix ls --here rendering undefined in the identity column for data rows, and add --json and revert --json dropping the key entirely; ls --here --json gains lockedSha, the name status --json already uses.
  • Stop show reporting (update available) for every item forever, including immediately after a clean add; it compared a content hash of the data repo's working tree against the locked pin digest, two schemes over two inputs that can never be equal. status owns update detection and reported these items correctly all along.
  • Stop status exiting 1 for a whole report when one subagent's locked tree resolves but cannot be read.
  • Stop add re-pinning after consent, so a data-repo commit landing in that window can no longer add a target the consent gate never saw.
  • Correct three user-facing strings: --kind help omitted subagents, the lock description promised an inspect command, and the "data repo not found" error named a default path capshelf never reads.