When the binary lived in a root-owned directory, upgrade re-ran the whole
command under sudo. The built-in skill download, UI asset cache, and oplog
then landed in the user's home as root-owned files, and a later git pull
of the skills source failed with "unable to unlink old
'skillshare/SKILL.md': Permission denied".
The sudo child now only receives an internal --replace-binary=<version>
flag that downloads and swaps the binary. The parent waits for it and
does everything else as the invoking user. The needs-password hint shown
by the CLI and dashboard drops its sudo prefix, since a fully root
upgrade reproduces the same breakage.
The 0.21.9 entry was written before the toolbar was compacted and before
disabled tree items got their own icon, and its folder bullet still
described Group as a separate dropdown.
Six labelled selects, each reading "All" most of the time, plus a
two-button expand/collapse group made the toolbar wrap and hard to scan.
- Filters are chips: an unset one shows only its name; a set one fills
in, shows its value and has a clear button.
- Group and sort, which change rarely, share one icon menu.
- Expand/collapse all is one button that flips with the tree's state.
The tree showed "Disabled" as hover text on each disabled skill, which
repeated what the dimmed row already said and crowded the hover slot.
A disabled item now swaps its icon for a power-off icon (with an
accessible name), and the hover text is kept only for folder counts.
Record the tree view rework, tracked-repo targets and folder grouping from #295, plus the dot-folder sync fix (#294), the Plugins page fix (#296) and the dashboard server crash fix.
Skills installed with --into land in folders, but the list and cards
views could only filter and group by source, so there was no way to
look at one folder at a time or see the folders side by side.
Add a Folder filter next to Targets and a Folder option to Group. A
skill's folder is its parent path, a tracked repo counts as one folder,
and skills at the source root group under "Root". The filter is hidden
when everything lives in one folder. Agents get the same controls.
Every API request reloads the skills and agents metadata stores and
swaps the server's pointers under s.mu, but several handlers read or
wrote the stores outside that lock: listing, skill detail, check, the
update stream, and the hash refresh after setting targets. Concurrent
requests could race on the pointer or on the stores' maps, which the
race detector reported hundreds of times under load and which can
crash the server with a concurrent map access.
Store access now always happens under s.mu. Reads return a copy of
the entry so it can be used after unlocking, and target handlers hash
files outside the lock and only write the result under it.
Turning a whole folder or tracked repo off meant ticking every skill
in it, and the tree rows mixed table columns with indentation.
The tree view is now an explorer tree beside a detail pane. Click a
folder, repo or skill to select it; Cmd/Ctrl-click adds, Shift-click
selects a range. The pane turns the whole selection on or off with one
switch, sets its targets (tracked repos included), and lists each skill
with its own switch. Counts and disabled markers show on hover, and the
divider between the two sides can be dragged.
Enable/disable toasts now replace each other through a new toast `key`,
so quick on/off clicks no longer leave an outdated "Enabled" on screen.
Refs #295
Setting targets on a skill inside a tracked repo used to be refused,
because writing its SKILL.md leaves the clone dirty and `update` then
skips the repo (or `--force` throws the setting away).
Store those targets in the source's .metadata.json instead, keyed by
the skill's path, and apply them at discovery over the frontmatter.
The clone stays clean, `update` keeps working, and the setting survives
it. Uninstalling a repo or skill drops its entries.
Overrides are saved in the same lock section that writes them: every
API request reloads the store from disk, so saving after unlocking
could lose them, and saving outside the lock raced with other writers.
Refs #295
When an Agent's plugin inventory failed (for example OpenCode with both
opencode.json and opencode.jsonc present), the inventory call overwrote
the host's installed list with nil, which the API encoded as null. The
Plugins page read its length unguarded, so one misconfigured Agent
crashed the whole page.
Host.fail now restores an empty list, covering every inventory path
that reports through it, and the dashboard reads a null list as empty
so a single Agent can no longer take the page down.
Refs #296
Flat target naming turns a source skill under a dot directory
(e.g. .system/example) into a hidden target entry (.system__example).
Target scans skipped every hidden name, so diff, doctor, status and the
dashboard reported these synced skills as missing, and prune never
removed them after the source skill was deleted.
Hidden entries are now skipped only when the target manifest does not
record them as managed, so files like .skillshare-manifest.json or
.DS_Store are still ignored.
Refs #294
The pages still told users to inspect conflicts with git status and to
fix diverged tracked repos with git pull --rebase, but pull now merges
diverged history, resolves .metadata.json conflicts on its own, and
undoes the merge on any other conflict, so those steps no longer match
what happens. Also documents the Git Sync page fetch, the Pull count,
and the push_rejected API error in every translated locale.
Agents follow these skills literally, and several no longer matched the
repo: the test example called AssertSuccess() without t so every test
written from it failed to compile, update-docs grepped for a flag package
the CLI never uses, codebase-audit checked a targets.yaml schema and a
specs/ directory that are gone, and the auth-test recipe pointed at the
wrong package and ran credential-helper outside the login shell that puts
it on PATH. Also reconciles rules that contradicted AGENTS.md (host-side
test runs, unconditional scope confirmation, commit/tag authorization)
and drops a reference to an mdproof skill that is not installed.
Add a shared instruction kernel and byte-budgeted topic router so agents load only task-relevant repository guidance. Wire drift checks into local and CI workflows and point project skills at the routed topics.
Two machines sharing a source repo kept stopping at the same two places.
The Git page had no way to know the remote moved until a push bounced, so
it now fetches once on open and labels Pull with the commit count (status
itself stays offline because the sidebar polls it). And both machines
rewrite .metadata.json on every install or update, so nearly every
diverged pull conflicted there; those conflicts are now merged per entry,
keeping the later installed_at, while any other conflict still stops the
pull and names the files.
Once both machines committed, the UI push was rejected and the follow-up
pull failed with "Need to specify how to reconcile divergent branches",
leaving no way out without a terminal. Pull now merges explicitly and
aborts a conflicted merge so a later commit-all cannot push conflict
markers; a rejected push returns push_rejected so the UI offers Pull.
In copy mode, sync --force can push source over target, but collect
skipped every file already in source, so edits made directly in a
copy-mode target could not be pulled back. collect --force now
overwrites those files; identical ones are still skipped.
Collect also replaced collected files with symlinks regardless of mode,
turning copy-mode targets into symlinks. Copy-mode targets now keep
their files, in the CLI and the extras TUI.
With flatten, two target files sharing a basename map to one source
file; the second is now reported instead of overwriting the first under
--force. --from paths are matched after expanding ~ and cleaning, so a
spelling that differs from config no longer drops the target's mode.
Refs #291
The Targets row and the -p mode notes did not mention that a target
now lists and switches its MCP servers, nor that its Sync writes every
target's MCP files.
A target's page covered its skills and agents only, so seeing which MCP
servers an Agent gets meant reading across the MCP page's matrix. A
target with an MCP config file now has an MCP tab, in global and -p
mode, with a row per server that saves as it is clicked, and the
Targets list counts those servers beside the skills.
MCP sync writes every target's files at once, so the tab names its
button for that. Its confirm dialog, like the MCP page's and the
project sync dialog, now lists one row per Agent file the way the
Skills sync dialog does, and still says why an MCP entry conflicts.
Each save previews first and sends that revision, so a second click
racing it was refused and rolled back the first; the toggles now wait.
A switch that names no targets no longer gets saved with Agents that
cannot turn a server off, which the backend refused.
The Pi adapter settings moved from a hover icon to a line on the server
row, and Import from an Agent on a project's conflict now reads that
project's file, so the dashboard descriptions no longer matched.
Refs #289
Refs #290
Direct tools and other adapter settings were only visible on hover, so
checking which tools a server exposes to Pi meant pointing at every row.
They now sit on a line under the endpoint. Other adapter settings only
say that they are set, since they may hold anything.
Refs #289
Import from an Agent on a project's conflict read the Agent's global
file and wrote to the global source, so the project's own entry could
never be taken over from there. It now reads that project's file and
imports into the project.
The MCP page also crashed on a cold load: a hook ran after the loading
return, so the first render with data called one more hook than before.
Refs #290
Servers added to an Agent directly never showed up in the dashboard, so
they could only be found by opening Import for each Agent in turn. The
MCP page and each project's MCP tab now note which Agent files hold
servers skillshare does not manage, and Import opens on that file; a
project's import reads the file in that project's folder.
Refs #290
Removing a server always ended with its entries cleared from the Agent
config files on the next sync, so there was no way to hand a server back
to the Agent. Stop managing drops it from the source and forgets which
entries it wrote; the files stay as they are and sync leaves them alone.
Available as `mcp remove --keep-files`, in the remove wizard and as a
third choice in the dashboard's remove dialog.
Refs #290
Global init/UI error messages were indistinguishable from project-scope
ones and didn't say "global", so seeing "already initialized" from
`skillshare init` and "source directory not found" from `ss ui -g`
looked contradictory. The latter's suggested fix ('skillshare init')
also led nowhere, since it just hits the already-initialized check
without recreating the missing source directory.
The Skills and Agents lists count project targets in their icon stack,
but the detail page showed only global targets, so the two disagreed.
The detail page now lists each project the resource syncs to and links
to it, since project filters are set per project, not per target.
The Skills and Agents sync dialogs previewed with a dry-run, which reports
every existing link as linked again, so they showed changes on targets
that were already in sync and disagreed with the Sync page and the
project dialog. They now read the same diff as the pending dots.
Refs #289
A project's pending changes could only be written by a full sync, which
also touched global targets and every other project. Sync project on the
project page previews and writes just that project's skills, agents and
MCP; the API takes a project root to scope the resource sync and the MCP
apply.
Refs #289
Syncing meant leaving the page for Sync and reviewing every part at once.
Skills, Agents and MCP now each have a sync button that previews and
writes only that part in a dialog, and the Go to Sync spots after an
update, uninstall or collect open the same dialog. Pi adapter settings
are shown on hover so they can be checked without opening Edit.
Refs #289
Unticking every Agent dropped directTools and piOptions from the source,
so ticking Pi again came back with the adapter reset. The settings now
stay until the adapter itself is changed.
Refs #289
The available-target list comes from a Go map, so the Agents in the
add-target dialog's account mode changed order on every load. Sort them
by name like the known-target list, which puts claude first and pi last.
Refs #289
A target declared with `agent` and `config_dir` is another config directory of a
built-in Agent, such as a second Claude account. Its skills and agents paths are
derived from the Agent's defaults at load time, so the config only records the
directory. The same target name works for MCP (written in the Agent's format into
the account's own file), for plugins (the Agent CLI runs with CLAUDE_CONFIG_DIR,
CODEX_HOME or PI_CODING_AGENT_DIR) and for `mcp import --from`. Removing such a
target warns when MCP config still names it instead of failing later at sync.
The dashboard gains an "another account" mode in the add-target dialog and lists
accounts next to the Agents on the MCP and plugin pages.
Refs #289
The MCP row on the ui page kept English button names in the translated pages,
next to labels that were already translated, and the English page itself said
Import where the button reads Import from a target. Three locales also lacked
the paragraph on upgrading without a terminal to ask for a password on.
The dashboard pages still described the project row without the Agent logos
and the Match the project action, the ui page did not mention where piOptions
is edited, and upgrade did not say the download now reports its progress.
Refs #289
The row listed the native field written per Agent in a narrow right-aligned
column, which was tiring to read and meant nothing to someone who has not seen
those files. It now shows the Agents' logos, keeps a sentence only for what
needs a reason, and an old entry with its own Agent list gets a one-click way
to follow the project instead of an instruction.
Refs #289
The dialog rebuilds a server from its fields, so without one for piOptions any
edit from the dashboard would drop what the config holds. It reuses the snippet
editor and refuses text that is not a JSON object, or that sets a field
Skillshare writes itself, before anything is sent.
Refs #289
The adapter has per-server fields such as excludeTools and approveTools that
Skillshare has no setting for, so they could only live in Pi's file, on one
machine. piOptions carries them as given. Like directTools they stay out of
the ownership hash, so a value someone already added by hand is not a conflict.
Refs #289
A project's off switch that was edited by hand in the Agent's file shows
up as a conflict with "Import from Agent" next to it. Import reads the
Agent's global file, which holds the server and never the project's
switch, so it brought in something unrelated and the conflict stayed.
Only replacing the entry settles it, so that is the one button offered.
The conflict row also says which project folder it is in: one server
can conflict in several, and two such rows shared a React key.
Refs #289
Pi reads one config file per project through one extension. A switch
took pi-mcp-adapter from the global server, so in a project whose own
servers use pi-mcp-extension it put two extensions into one file and the
whole plan failed. From the dashboard that was an error toast on
flipping a switch that had nothing to do with Pi.
Pi has no per-project switch with pi-mcp-extension, so the switch now
skips Pi there and still reaches the other Agents. The project page
says the same: Pi is listed among the Agents that keep loading it.
Refs #289