Add per-agent scoping to skill add --mcp - #522
Merged
Merged
Conversation
xsh310
force-pushed
the
xshen/skill-per-agent-add
branch
from
September 8, 2026 05:43
3e8757f to
a3378e8
Compare
xsh310
force-pushed
the
xshen/skill-per-agent-add
branch
from
September 8, 2026 18:25
a3378e8 to
4918b2d
Compare
xsh310
force-pushed
the
xshen/skill-per-agent-add
branch
from
September 8, 2026 19:09
4918b2d to
b1219a3
Compare
xsh310
marked this pull request as ready for review
September 8, 2026 20:31
xsh310
force-pushed
the
xshen/skill-per-agent-add
branch
2 times, most recently
from
September 8, 2026 22:16
fdeee67 to
8589f8d
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
sunishsheth2009
approved these changes
Sep 11, 2026
xsh310
force-pushed
the
xshen/skill-per-agent-add
branch
from
September 11, 2026 03:07
879247f to
268881e
Compare
Give `ucode skill add --mcp` an `--agents` option so a schema can be added to a chosen subset of configured coding agents instead of all of them, mirroring `ucode mcp add --agents`. `add_skills_command` takes an optional `agents` set and forwards it to `setup_mcp_clients`, which scopes the client set; the per-client map is updated only for the targeted clients. `--agents` is rejected outside `--mcp` since downloaded skills use shared directory families. The `--mcp --agents` path bootstraps only agents that are not configured for MCP yet, so re-targeting an already-configured agent no longer re-runs the agent setup (re-login, binary reinstall, re-validate). It computes the not-yet-ready subset and passes just those to `_configure_agents_for_mcp`, while still scoping the skills update to every named agent. Co-authored-by: Arthur Jenoudet <arthur.jenoudet@databricks.com> Co-authored-by: Isaac <no-reply@databricks.com>
`skill add --agents ""` (or `,`) raised a hard "No agents provided" error, while the other agent-scoped commands (`mcp add`, `mcp remove`) fold an empty --agents value to None and act globally. Match them so `skill add --mcp` mirrors `ucode mcp add` exactly: an empty --agents now targets every configured agent instead of erroring. Co-authored-by: Arthur Jenoudet <arthur.jenoudet@databricks.com> Co-authored-by: Isaac <no-reply@databricks.com>
Push the skip-already-configured logic into _configure_agents_for_mcp and have it return the full canonical scope, so `ug skill add --mcp --agents` and `ug mcp add --agents` share one path. skills_add no longer reimplements agent normalization, the cursor special-case, or the bootstrap loop; mcp add now also skips reconfiguring already-set-up agents. Co-authored-by: Isaac <no-reply@databricks.com>
xsh310
force-pushed
the
xshen/skill-per-agent-add
branch
from
September 11, 2026 04:25
268881e to
caa792e
Compare
xsh310
added a commit
that referenced
this pull request
Sep 11, 2026
## What `ug status` rendered a single `Skill MCP Locations` line from the flat scope mirror. Once agents carry different skill scopes (which per-agent add now makes possible), that single line hides the divergence. This makes `status` report each agent's real scope. ## How - The Skills section reads each configured client's scope with `skill_locations_for_client`. - When every agent shares the same scope, it collapses to the original single `Skill MCP Locations` + `Configured` lines. - When scopes diverge, it prints a per-agent line (`Claude Code skill MCP locations: ...`, `Codex skill MCP locations: ...`), matching how `_print_skills_summary` already reports scopes. ## Tests - `test_cli.py`: a state whose per-client scopes diverge renders one line per agent with that agent's own locations. The existing uniform-scope and no-locations status tests still pass unchanged. `uv run pytest tests/test_mcp.py tests/test_cli.py tests/test_lint.py` is green. ### Manual verification (installed build) Using the sandbox from the per-agent add PR (agents scoped to divergent locations: claude `shared.skills, claude.only`, codex `shared.skills`), ran `ucode status`. The Skills section rendered one line per agent: ``` Skills Claude Code skill MCP locations: shared.skills, claude.only Codex skill MCP locations: shared.skills ``` The new `ug skill add` / `ug skill remove --mcp` hint also appeared. The uniform-scope collapse path is covered by the existing status tests. ## Stacking Second in the stacked per-agent skills series, based on `xshen/skill-per-agent-add` (#522). 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 an
--agentsoption toucode skill add --mcpso a skill schema can be added to a chosen subset of configured coding agents instead of all of them. This mirrorsucode mcp add --agents, closing the gap between how the skills MCP path and the general MCP path scope work per agent.How
add_skills_command(locations, agents=None)forwardsagentstosetup_mcp_clients, which scopes the returned client set. The per-client skill-location map is updated only for the targeted clients; the rest keep their scope untouched.cli.py,skill addgains the--agentsoption, bootstraps any named agent that is not configured yet via_configure_agents_for_mcp, and guards that--agentsis only valid with--mcp(downloaded skills use shared directory families and cannot be agent scoped).--agents, behavior is unchanged: every configured agent is updated.Tests
test_mcp.py: per-agent add touches only the named client's scope and re-registers only that client; a global add reaches every configured client while preserving per-client divergence; adding a schema an agent already has is a no-op.test_cli.py:--agentsis parsed, bootstrapped, and forwarded for--mcp; an empty--agentslist is rejected;--agentsis rejected in download mode.uv run pytest tests/test_mcp.py tests/test_cli.py tests/test_lint.pyis green.Manual verification (installed build)
Installed the stack tip with
uv tool install --reinstall .(ucode0.1.0+91.g5c0dc1e) and exercised the command in an isolated sandbox: a temporaryHOME, stubclaude/codexbinaries that record every registration ucode issues, andDATABRICKS_BEARERset so auth is satisfied offline. State was seeded with both agents configured and no skills.ucode skill add --location shared.skills --mcpshared.skills; each re-registered onceadd_skills_command(["claude.only"], agents={"claude"})claude=[shared.skills, claude.only],codex=[shared.skills]; only claude re-registered; summary showed divergent per-client URLs--locationucode skill addERROR --location is required for ucode skill add.--agentswithout--mcpucode skill add --location a.b --agents claudeERROR --agents is only supported when using --mcp--agentsucode skill add --location a.b --mcp --agents ,ERROR No agents provided for --agents ...The
--agentsadd path first runs an interactive agent bootstrap, so the per-agent case droveadd_skills_command(..., agents=...)directly against the installed package with only that prompt stubbed; state I/O and per-client agent registration ran for real. CLI parsing and bootstrap are covered by the mocked unit tests.Stacking
This is the first of a stacked series that adds per-agent scoping to the skills MCP connection, built on top of
xshen/per-client-skill-scopes(per-client skill scope map) andskill-managed-download-only(managed skills delivered by download only). Reviewing the diff against its base branch 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.