Add skill remove --mcp for global scope removal - #524
Merged
Conversation
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 8, 2026 05:43
fb94ba5 to
59d2fd5
Compare
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 8, 2026 18:25
59d2fd5 to
0aa87df
Compare
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 8, 2026 19:09
0aa87df to
98934ea
Compare
xsh310
marked this pull request as ready for review
September 8, 2026 21:09
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 8, 2026 22:12
98934ea to
4921ec5
Compare
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 8, 2026 22:16
4921ec5 to
333fc68
Compare
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 8, 2026 23:09
14e8dcd to
bda1dc8
Compare
xsh310
removed this pull request from stack #521
September 10, 2026 21:22
xsh310
added this pull request to stack #556
September 10, 2026 21:22
xsh310
force-pushed
the
xshen/skill-remove
branch
2 times, most recently
from
September 11, 2026 04:25
915b95c to
86e4fc9
Compare
Introduce `ucode skill remove --mcp`, the inverse of `ucode skill add --mcp`. It shows the schemas in each configured client's skills scope, annotated with the clients that carry them, and removes the selected schemas from every client that has them. `remove_skills_command` reuses the per-client update path, so dropping the last schema leaves the schema-less connection in place. It never adds or reconfigures anything and needs no Databricks auth. Removing downloaded skills is not supported yet, so the command requires `--mcp`. A status hint points at the new add/remove commands. Co-authored-by: Arthur Jenoudet <arthur.jenoudet@databricks.com> Co-authored-by: Isaac <no-reply@databricks.com>
remove_skills_command called _update_skills_mcp with the default print_summary=True, so a removal printed the full "Skills MCP registered" block (Server / URL / Configured / Tools) before "Removed N skill schemas", which reads wrong for a removal. Pass print_summary=False so a removal reports only what it removed. The print_summary hook already existed for this case. Co-authored-by: Arthur Jenoudet <arthur.jenoudet@databricks.com> Co-authored-by: Isaac <no-reply@databricks.com>
xsh310
force-pushed
the
xshen/skill-remove
branch
from
September 11, 2026 05:32
86e4fc9 to
a1a5e49
Compare
sunishsheth2009
approved these changes
Sep 11, 2026
xsh310
added a commit
that referenced
this pull request
Sep 11, 2026
## What
Adds an `--agents` option to `ucode skill remove --mcp` so a schema can
be removed from a chosen subset of configured agents and kept on the
rest, mirroring `ucode mcp remove --agents`. This is the core bug fix in
the series.
## The bug
Before per-agent scopes, an all-agent add wrote a shared layer while a
per-agent remove edited a separate one, so they did not compose: `skill
add --mcp X` (all) followed by `skill remove --mcp --agents claude` was
a silent no-op, and often printed a misleading "nothing to remove". Now
that developer scope is a single per-agent map, the two operations act
on the same layer.
## How
- `remove_skills_command(agents=None)` forwards `agents` to
`setup_mcp_clients`, which scopes the client set. The picker offers only
those clients' schemas and removal edits only their maps; other agents
keep the schema and are not re-registered.
- The "nothing to remove" note names the scope when `--agents` was
given.
- `cli.py`: `skill remove` gains the `--agents` option; without it,
removal stays global.
## Tests
- `test_mcp.py`: add-all then `remove --agents claude` removes the
schema for claude only and re-registers only claude; the picker is
offered only the named clients' scope; an empty named scope is a no-op.
- `test_cli.py`: `--mcp` forwards `agents=None`; `--mcp --agents
claude,codex` forwards the parsed set.
`uv run pytest tests/test_mcp.py tests/test_cli.py tests/test_lint.py`
is green.
### Manual verification (installed build)
This is the headline fix, so I verified the exact previously-failing
sequence end to end. In the sandbox (installed build
`0.1.0+91.g5c0dc1e`, stub `claude`/`codex` binaries recording
registrations, offline auth via `DATABRICKS_BEARER`), `shared.skills`
had been added to both agents. Then I invoked
`remove_skills_command(agents={"claude"})` with the picker stubbed to
select `shared.skills`:
| Check | Result |
|---|---|
| Picker input | Offered only the named client's scope:
`{claude:[shared.skills, claude.only]}` |
| State after | `shared.skills` removed from claude only:
`{claude:[claude.only], codex:[shared.skills]}` |
| Registrations | Only claude re-registered; codex untouched |
Before this change the same sequence was a silent no-op (add wrote a
shared layer, per-agent remove edited a different one). It now removes
for the named agent while the other keeps the schema.
## Stacking
Fourth in the stacked per-agent skills series, based on
`xshen/skill-remove` (#524). Reviewing the diff against that base shows
just this change. It rebuilds behavior originally designed by Arthur
Jenoudet on the current per-client-map state model.
This pull request and its description were written by Isaac.
Co-authored-by: Arthur Jenoudet <arthur.jenoudet@databricks.com>
Co-authored-by: Isaac <no-reply@databricks.com>
xsh310
added a commit
that referenced
this pull request
Sep 11, 2026
## What Documents the per-agent skills behavior this stack introduced but the README did not yet cover: - `--agents` scoping on `ug skill add --mcp` (add schemas to specific agents; any not set up yet are configured first). - `ug skill remove --mcp`, with `--agents` to scope removal to named agents. - Each agent's skill MCP scope now shown by `ug status`. ## How - Added a "Remove skill scopes" subsection, plus an `--agents` example and note in "Add skill scopes". - Added quick-reference rows for the per-agent `skill add` and both `skill remove` forms. - Updated the `ug status` row to mention per-agent skill MCP scope. - Aligned the existing `skill add` examples from `ucode` to `ug` for consistency with the rest of the README. Docs only; no code changes. ## Stacking Stacked on `xshen/skill-per-agent-remove` (#525), the top of the per-agent skills series. It documents the user-facing surface added across #520, #522, #523, #524, and #525. This pull request and its description were written by Isaac. Co-authored-by: Isaac <no-reply@databricks.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds
ucode skill remove --mcp, the inverse ofucode skill add --mcp. It interactively removes skill schemas from the skills MCP connection across every configured agent. Removing downloaded skills is not supported yet, so the command requires--mcp.How
remove_skills_command()reads each configured client's skills scope, offers the union of schemas via a picker (each annotated with the clients that carry it), and removes the selected schemas from every client that has them.cli.pygains theskill removecommand and astatushint pointing atug skill add/ug skill remove --mcp.Tests
test_mcp.py: a selected schema is removed from every client and each is re-registered; removing all schemas keeps the schema-less connection; an empty scope is a no-op that never opens the picker; the picker is offered each client's real scope.test_cli.py:skill removewithout--mcperrors;skill remove --mcpdispatches the global removal.uv run pytest tests/test_mcp.py tests/test_cli.py tests/test_lint.pyis green.Manual verification (installed build)
Ran against the installed build (
0.1.0+91.g5c0dc1e) in the isolated sandbox (tempHOME, stubclaude/codexbinaries recording registrations,DATABRICKS_BEARERfor offline auth), starting from stateclaude=[claude.only],codex=[shared.skills]. Invokedremove_skills_command()with the interactive picker stubbed to selectshared.skills(the picker needs a TTY); all state I/O and per-client registration ran for real.{claude:[claude.only], codex:[shared.skills]}shared.skillsdropped from codex; result{claude:[claude.only]}with codex's now-empty entry removeducode skill remove(no--mcp) exits withERROR Removing downloaded skills is not supported yet ...Stacking
Third in the stacked per-agent skills series, based on
xshen/skill-per-agent-status(#523). Per-agent remove builds on this next. Reviewing the diff against that base shows just this change. It rebuilds behavior originally designed by Arthur Jenoudet on the current per-client-map state model.This pull request and its description were written by Isaac.