fix(mcp): rewrite skill-shipped MCP command paths for the Claude target by srid · Pull Request #1465 · microsoft/apm

srid · 2026-05-23T23:07:54Z

Summary

Self-defined MCP launchers shipped inside a skill package (e.g. juspay/nix-chrome-devtools-mcp) declare their command: using the cross-tool converged skill path .agents/skills/<name>/.... Every other target deploys skills to .agents/skills/, so the upstream value is correct as-declared — but Claude is the documented outlier (see install/skill_path_migration.py:55: ".claude/skills/ is excluded (Claude is not in the convergence set)"). Consumers whose apm.yml has target: claude end up with files at .claude/skills/<name>/... while .mcp.json still points at the non-existent .agents/skills/<name>/..., so Claude Code fails to launch the server.

Repro

# apm.yml
target: claude
dependencies:
  apm:
    - juspay/nix-chrome-devtools-mcp

$ apm install
$ ls .agents/skills/nix-chrome-devtools-mcp/bin/serve
ls: cannot access '.agents/skills/...': No such file or directory
$ ls .claude/skills/nix-chrome-devtools-mcp/bin/serve   # the file IS here
-rwxr-xr-x ...
$ jq -r '.mcpServers["chrome-devtools"].command' .mcp.json
.agents/skills/nix-chrome-devtools-mcp/bin/serve   # ← wrong

Fix

Override _format_server_config in ClaudeClientAdapter to rewrite the .agents/skills/ prefix to .claude/skills/ on the way out. Scoped to the prefix only — bare binaries (npx, node), absolute paths, and unrelated relative paths pass through unchanged.

Drift detection (_detect_mcp_config_drift) compares the upstream dep.to_dict() against the lockfile baseline (both upstream-shaped), not against on-disk paths, so the rewrite does not perturb idempotence. Verified: a second apm install reports "already configured" and the on-disk path remains rewritten.

Why this layer

_build_self_defined_info in mcp_integrator.py builds one synthetic_info per dep that is reused across all target_runtimes — rewriting there would either need per-runtime copies or runtime-specific knowledge in the integrator.
Each adapter already owns its target-specific shaping (cf. _normalize_mcp_entry_for_claude_code), and Claude is the only target that needs this rewrite, so a single override stays local.

Test plan

New unit test: .agents/skills/<name>/... → .claude/skills/<name>/... end-to-end through configure_mcp_server → .mcp.json
New unit test: pass-through for npx, absolute paths, ./scripts/..., .agents/other/...
Full tests/unit/test_claude_mcp.py suite: 28 passed (+4 subtests)
Adapter regression: tests/unit/adapters, tests/unit/test_copilot_adapter.py, tests/integration/test_claude_mcp_schema_fidelity.py: 116 passed
Integration regression: tests/unit/integration, tests/integration/test_agent_skills_target.py: 1444 passed
Manual E2E: re-ran apm install in a target: claude project consuming juspay/nix-chrome-devtools-mcp; .mcp.json now points at the deployed launcher and ./.claude/skills/nix-chrome-devtools-mcp/bin/serve boots successfully

Skill packages that ship a self-defined MCP launcher (e.g. juspay/nix-chrome-devtools-mcp) declare their `command:` using the cross-tool converged path `.agents/skills/<name>/...`. Every target except `claude` deploys skills to `.agents/skills/`, so the upstream path is correct as-declared. Claude is the documented outlier -- `install/skill_path_migration.py:55` explicitly excludes `.claude/skills/` from the convergence -- and consumers with `target: claude` end up with files at `.claude/skills/<name>/...` while `.mcp.json` still points at the non-existent `.agents/skills/<name>/...`. Override `_format_server_config` in `ClaudeClientAdapter` to rewrite the `.agents/skills/` prefix to `.claude/skills/` before the entry is written. The rewrite is scoped to the prefix only (bare binaries like `npx`, absolute paths, and unrelated relative paths pass through) and runs once per format, so drift detection (which compares upstream `dep.to_dict()` against the stored baseline, not on-disk paths) is unaffected and reinstalls remain idempotent. Repro: `apm install` in a project with `target: claude` that depends on `juspay/nix-chrome-devtools-mcp` writes a `.mcp.json` whose `command` points at `.agents/skills/nix-chrome-devtools-mcp/bin/serve` -- a path that does not exist, so Claude Code fails to launch the server.

Copilot

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Updates the Claude client adapter to rewrite self-defined MCP skill launcher paths from the cross-tool .agents/skills/ convention to Claude’s .claude/skills/ location so generated .mcp.json entries point to paths that exist after deployment.

Changes:

Add command-prefix rewrite logic in ClaudeClientAdapter and apply it when formatting server config.
Add unit tests verifying rewrite behavior for skill-path commands and no-op behavior for other command shapes.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
tests/unit/test_claude_mcp.py	Adds regression tests covering `.agents/skills/` → `.claude/skills/` rewriting and ensuring non-skill commands are unchanged.
src/apm_cli/adapters/client/claude.py	Implements the command-prefix rewrite and applies it to formatted server configs for Claude.

srid · 2026-05-23T23:19:19Z

+    @classmethod
+    def _rewrite_self_defined_skill_command(cls, command: str) -> str:
+        if isinstance(command, str) and command.startswith(cls._CONVERGED_SKILL_PREFIX):
+            return cls._CLAUDE_SKILL_PREFIX + command[len(cls._CONVERGED_SKILL_PREFIX) :]


This is ruff format's style (Black-style, E203) for slices whose index is a complex expression — and the project's configured formatter, per pyproject.toml. Both ruff check src/apm_cli/adapters/client/claude.py and ruff format --check pass; removing the space would in fact be re-added by ruff format.

It's also the dominant convention in src/apm_cli/ — a quick grep:

src/apm_cli/bundle/plugin_exporter.py:72: normalized = normalized[len("skills/") :].lstrip("/") src/apm_cli/utils/normalization.py:41: return content[len(_BOM) :] src/apm_cli/core/conflict_detector.py:115: server_name = raw_key[len("mcp_servers.") :] src/apm_cli/commands/policy.py:54: return source[len(prefix) :] src/apm_cli/bundle/lockfile_enrichment.py:153: mapped = dst_prefix + f[len(src_prefix) :] src/apm_cli/core/command_logger.py:693: reason = reason[len(prefix) :] src/apm_cli/bundle/local_bundle.py:281: return value[len("sha256:") :].strip().lower() src/apm_cli/commands/marketplace/check.py:60: tag_name = tag_name[len("refs/tags/") :] src/apm_cli/core/token_manager.py:242: token = line[len("password=") :]

Leaving as-is.

## Summary `.mcp.json` shipped at HEAD points the `chrome-devtools` server at `.agents/skills/nix-chrome-devtools-mcp/bin/serve` — a path that **does not exist** on disk for this project. The launcher actually lives at `.claude/skills/nix-chrome-devtools-mcp/bin/serve` because our `apm.yml` declares `target: claude`, and Claude is the only apm target that does not converge to `.agents/skills/`. Result: Claude Code silently fails to start the chrome-devtools MCP server. ## What changed - **`.mcp.json`**: rewrite the `command` prefix from `.agents/skills/` to `.claude/skills/`. - **`apm.yml`**: add a comment explaining the bug and linking the upstream fix. ## Upstream fix [microsoft/apm#1465](microsoft/apm#1465) — `ClaudeClientAdapter` learns to rewrite the prefix when emitting `.mcp.json`. Once that ships in an apm release, `apm install` regenerates `.mcp.json` identically to the version committed here, so this patch becomes a no-op. ## Test plan - [x] Verified `./.claude/skills/nix-chrome-devtools-mcp/bin/serve` exists and starts cleanly - [x] `apm install` with the patched apm-cli produces the same `.mcp.json` as this PR (idempotent) - [ ] In Claude Code: `chrome-devtools` MCP server connects and exposes its tools

github-actions · 2026-05-24T02:54:26Z

APM Review Panel: `ship_with_followups`

Fix silent Claude MCP failure: rewrite .agents/skills/ command prefix to .claude/skills/ in ClaudeClientAdapter so self-defined skill packages resolve correctly at runtime.

cc @sergio-sisternes-epam -- a fresh advisory pass is ready for your review.

All six active panelists converge: the fix is correct, well-scoped, and introduces no regressions. python-architect confirms the Template Method override is the right pattern. supply-chain-security confirms no new path-traversal or credential risk. cli-logging-expert confirms the silent rewrite is appropriate for the happy path. The fix closes a real user-facing breakage -- Claude users installing skill packages with self-defined MCP launchers received a broken .mcp.json that failed silently only when Claude tried to start the server.

The highest-convergence signal is the missing CHANGELOG entry: both oss-growth-hacker and doc-writer flag it independently. This is not a paperwork concern -- a bug fix that affects Claude users but carries no CHANGELOG entry is a trust deficit. Users who hit the broken command path and worked around it manually will never know it was resolved; prospective contributors will not see evidence of responsive maintenance. A single line under [Unreleased] ### Fixed converts invisible maintenance into a community trust signal.

The test-coverage-expert finding is real but proportionate: unit tests cover the rewrite logic in isolation and the two passing tests are load-bearing evidence that the fix works. The missing integration-tier fixture test is a gap worth tracking -- if _format_server_config super-call or prefix constants drift, only unit coverage would catch it -- but it is not a reason to hold the PR. The devx-ux install-time warning suggestion is a genuine UX improvement and should be tracked as a follow-up issue, not a pre-merge requirement.

Aligned with: Multi-harness / multi-host -- Directly advances Claude harness fidelity: skill MCP launchers now resolve to the correct runtime path for Claude users. Portable by manifest -- Ensures the manifest-driven skill install produces a correct on-disk artifact (.mcp.json) across harnesses. Pragmatic as npm -- Silent runtime failure is the worst failure mode for a package manager; this fix makes Claude skill installs reliably functional without user intervention.

Growth signal. Claude is a high-visibility harness. A broken .mcp.json on skill install is a first-impression failure that converts curious Claude users into churned users. This fix, with a proper CHANGELOG entry, becomes a concrete signal that APM ships responsive, harness-aware fixes -- exactly the narrative that builds contributor and enterprise trust.

Panel summary

Persona	R	N	Takeaway
Python Architect	0	1	Minimal, well-scoped fix: Template Method override + isolated classmethod rewriter; only args-scope omission warrants a doc comment.
CLI Logging Expert	0	0	No CLI output changes; silent path rewrite is correct -- existing success/error surface is unchanged and appropriate.
DevX UX Expert	1	0	Fix is correct and invisible to users on the happy path; the only UX gap worth addressing is surfacing the broken-command-path failure at install time rather than silently deferring it to Claude runtime.
Supply Chain Security Expert	0	1	No supply-chain security regressions; the string-prefix rewrite is type-guarded, scope-limited, and does not amplify any pre-existing path or credential risk.
OSS Growth Hacker	1	0	Solid silent-failure fix for Claude users; needs a CHANGELOG entry so it converts to trust signal rather than invisible maintenance.
Doc Writer	1	0	Add a CHANGELOG [Unreleased] ### Fixed entry -- this bug fix affects Claude users and the section already exists but has no entry for this PR.
Test Coverage Expert	1	0	Unit tests are well-targeted and cover the rewrite logic in isolation; no integration-tier test verifies the on-disk .mcp.json command path after a real Claude skill install -- a recommended gap against the install-pipeline floor.

B = blocking-severity findings, R = recommended, N = nits.
Counts are signal strength, not gates. The maintainer ships.

Top 5 follow-ups

[OSS Growth Hacker + Doc Writer] Add CHANGELOG [Unreleased] ### Fixed entry for the Claude skill path rewrite. -- Dual-panelist convergence is the highest signal in this review. The fix is invisible to users without a changelog entry; with one, it becomes evidence of responsive maintenance and harness-aware quality. Suggested line: Fix Claude skill MCP launcher path: rewrite .agents/skills/ command prefix to .claude/skills/ in ClaudeClientAdapter so self-defined skill packages resolve correctly for Claude users. (#1465)
[Test Coverage Expert] Add integration-with-fixtures test: verify .claude/skills/ path in written .mcp.json after a real Claude skill install. -- Unit tests cover the rewrite logic but the install-pipeline surface has no integration-tier assertion. If prefix constants or the super-call drift, no automated check catches it before a user ships a broken .mcp.json. Track as a follow-up issue; suggested file: tests/integration/test_claude_mcp_schema_fidelity.py.
[DevX UX Expert] Add install-time existence check: emit _rich_warning if the rewritten command path does not exist on disk. -- Currently the broken-path failure is deferred to Claude runtime. A stat check after rewrite would surface the error at apm install time, aligning with the pragmatic-as-npm principle.
[Python Architect] Add a one-line docstring to _rewrite_self_defined_skill_command clarifying that only command is rewritten, not args. -- The intentional scope (command only) is not self-evident; without a comment a future contributor may file a follow-up bug or incorrectly extend the rewrite to args.
[Supply Chain Security Expert] Track path validation at the call-site as a future hardening task. -- The prefix rewrite preserves any traversal sequences already present in the command. Not a regression, but a resolved-path validation would close the theoretical risk surface.

Architecture

classDiagram
    direction LR
    class MCPClientAdapter {
        <<ABC>>
        +configure_mcp_server(name) bool
        +_format_server_config(info, env, vars) dict
    }
    class CopilotClientAdapter {
        <<ConcreteBase>>
        +_format_server_config(info, env, vars) dict
    }
    class ClaudeClientAdapter {
        <<ConcreteStrategy>>
        +_CONVERGED_SKILL_PREFIX str
        +_CLAUDE_SKILL_PREFIX str
        +_rewrite_self_defined_skill_command(cmd) str
        +_format_server_config(info, env, vars) dict
    }
    class CursorClientAdapter {
        <<ConcreteStrategy>>
        +_format_server_config(info, env, vars) dict
    }
    class GeminiClientAdapter {
        <<ConcreteStrategy>>
    }
    class WindsurfClientAdapter {
        <<ConcreteStrategy>>
    }
    MCPClientAdapter <|-- CopilotClientAdapter
    CopilotClientAdapter <|-- ClaudeClientAdapter
    CopilotClientAdapter <|-- CursorClientAdapter
    CopilotClientAdapter <|-- GeminiClientAdapter
    CopilotClientAdapter <|-- WindsurfClientAdapter
    note for ClaudeClientAdapter "Template Method: super()._format_server_config then rewrites .agents/skills/ -> .claude/skills/"
    class ClaudeClientAdapter:::touched
    classDef touched fill:#fff3b0,stroke:#d47600

flowchart TD
    A["configure_mcp_server (claude.py:249)"] --> B["_format_server_config (claude.py:67)"]
    B --> C["super()._format_server_config (copilot.py:408)"]
    C --> D["reads _raw_stdio.command (copilot.py:433)"]
    D --> E{"command starts with .agents/skills/?"}
    E -- yes --> F["_rewrite_self_defined_skill_command (claude.py:62) returns .claude/skills/... path"]
    E -- no --> G["command unchanged"]
    F --> H["config dict with rewritten command"]
    G --> H
    H --> I["FS write .mcp.json"]

Recommendation

Ship after adding the CHANGELOG entry (highest-convergence follow-up, one line, pre-merge preferred). All other follow-ups are post-merge: integration fixture test, install-time warning, docstring, and path validation. The fix is correct, the unit evidence is solid, and the only pre-merge gap is the missing changelog line that converts this invisible maintenance into a community trust signal.

Full per-persona findings

Python Architect

[nit] args not rewritten -- acceptable but worth a comment at src/apm_cli/adapters/client/claude.py:62
Only command is rewritten; args entries that might contain .agents/skills/ paths are left as-is. For the current skill launcher convention (command holds the binary path, args are flags) this is correct. A one-line docstring on _rewrite_self_defined_skill_command stating the intentional scope (command only, not args) would prevent a future contributor from filing a follow-up bug.
Suggested: Add docstring: "Rewrites the command field only; args are intentionally left unchanged as skill launchers place the binary path in command, not args."

CLI Logging Expert

No findings.

DevX UX Expert

[recommended] Silent runtime failure deferred from install to Claude launch at src/apm_cli/adapters/client/claude.py (_format_server_config)
The fix corrects a silent runtime failure -- broken command path in .mcp.json only surfaces when Claude tries (and fails) to start the MCP server, not at apm install time. Consider adding an install-time existence check (or at least a warning) when the resolved command path does not exist on disk.
Suggested: After rewriting the command, stat the resolved path and emit a _rich_warning if it does not exist. This turns a silent runtime break into a recoverable install-time prompt.

Supply Chain Security Expert

[nit] Path traversal sequences in command are preserved by the prefix rewrite at src/apm_cli/adapters/client/claude.py
The prefix rewrite preserves any path traversal sequences already present in the command (e.g. .agents/skills/../../../evil -> .claude/skills/../../../evil). This is not a regression introduced by this PR -- the risk existed before and is equal in both forms -- but it is a signal that the call-site should eventually validate the resolved path. No action required for this PR.

OSS Growth Hacker

[recommended] No CHANGELOG entry for a user-facing Claude bug fix at CHANGELOG.md
The [Unreleased] section has no entry for this fix. Claude users who hit the silent MCP failure would never know it was resolved -- and prospective Claude users evaluating APM won't see evidence that Claude first-run quality is actively maintained.
Suggested: Add a line under [Unreleased] ### Fixed: Fix: rewrite .agents/skills/ command prefix to .claude/skills/ in ClaudeClientAdapter so skill MCP launchers resolve correctly for Claude users. (#1465)

Doc Writer

[recommended] Missing CHANGELOG [Unreleased] entry for the Claude skill path rewrite fix at CHANGELOG.md
This PR fixes a user-facing bug: Claude users installing skill packages with self-defined MCP launchers received broken command paths (.agents/skills/ instead of .claude/skills/). The CHANGELOG [Unreleased] section has a ### Fixed block (PR fix: handle MCP v0.1 runtimeArguments.variables in copilot and codex adapters #1461) but no entry for this fix. Per project conventions, bug fixes that affect user-observable behavior belong in CHANGELOG [Unreleased] ### Fixed. The 0.14.0 release notes explicitly document the .claude/skills/ routing contract, so readers who encounter this breakage and search the changelog will find no record of the fix.
Suggested: Add under ## [Unreleased] ### Fixed: Fix Claude skill MCP launcher path: rewrite .agents/skills/ command prefix to .claude/skills/ in ClaudeClientAdapter so self-defined skill packages resolve correctly for Claude users. (#1465)

Test Coverage Expert

[recommended] Install-pipeline surface: no integration-tier test verifying .claude/skills/ path in written .mcp.json
Unit tests added are well-targeted but stay at unit tier. The tier-floor matrix requires integration-with-fixtures for the install-pipeline surface. No integration test in tests/integration/ covers the .agents/skills/ -> .claude/skills/ rewrite via a real fixture package. If _format_server_config super-call or prefix constants drift, unit tests catch the rewrite logic in isolation, but no integration-tier assertion would fail before a user ships a broken .mcp.json.
Proof (passed at unit): tests/unit/test_claude_mcp.py::test_self_defined_skill_command_rewritten_to_claude_skills -- proves: When a skill MCP launcher declares .agents/skills/... as its command, configure_mcp_server writes .claude/skills/... to .mcp.json [devx, multi-harness-support]
self.assertEqual(data['mcpServers']['chrome-devtools']['command'], '.claude/skills/nix-chrome-devtools-mcp/bin/serve')
Proof (passed at unit): tests/unit/test_claude_mcp.py::test_self_defined_non_skill_command_left_unchanged -- proves: Bare binaries, absolute paths, and unrelated relative paths are not rewritten [devx]
self.assertEqual(self.adapter._rewrite_self_defined_skill_command(raw_command), raw_command)
Proof (missing at integration-with-fixtures): tests/integration/test_claude_mcp_schema_fidelity.py::test_skill_mcp_launcher_command_rewritten_to_claude_skills_prefix -- proves: After apm install of a Claude-targeted skill with an MCP launcher, the on-disk .mcp.json has a .claude/skills/ command path [devx, multi-harness-support, portability-by-manifest]

Auth Expert -- inactive

PR only rewrites a filesystem command path prefix in ClaudeClientAdapter._format_server_config; no auth flows, token resolution, credential handling, or AuthResolver touched.

_{This panel is advisory. It does not block merge. Re-apply the

panel-review label after addressing feedback to re-run.}

Note

🔒 Integrity filter blocked 4 items

The following items were blocked because they don't meet the GitHub integrity level.

#1465 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
fix(mcp): rewrite skill-shipped MCP command paths for the Claude target #1465 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
3a7389d list_commits: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
#1465 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

Generated by PR Review Panel for issue #1465 · ● 2.6M · ◷

* chore: cut 0.15.0 Move Unreleased -> [0.15.0] - 2026-05-27 and bump pyproject + uv.lock. Audit applied: every PR merged since v0.14.2 has exactly one changelog entry; each entry leads with the user-visible impact. Fixes during audit: - Add missing entries for #1367, #1403, #1465, #1487, #1492, #1462, #1477, #1439, #1484, and the 131679f follow-up commit. - Collapse the two #1473 lines into one. - Merge the #1476 Security/GitCache-hardening entry into its Added entry (same PR, one logical change). - Replace bogus #1243 PR ref with the actual merge PR #1308 for the persisted transport-flag config. - Relocate the #1324-delivered marketplace CLI entries (apm pack --marketplace / --marketplace-path / --json, outputs map form) out of Unreleased and into [0.14.2], where they actually shipped. They were mis-attributed to #1317 and orphaned across the 0.14.2 cut. Verified locally: ruff check + ruff format --check both clean. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: danielmeppiel <danielmeppiel@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings May 23, 2026 23:07

srid requested review from danielmeppiel and sergio-sisternes-epam as code owners May 23, 2026 23:07

Copilot AI reviewed May 23, 2026

View reviewed changes

sergio-sisternes-epam added the status/needs-triage New, awaiting maintainer review. label May 23, 2026

srid mentioned this pull request May 23, 2026

Fix .mcp.json path for the chrome-devtools MCP server srid/anywhen#25

Merged

3 tasks

sergio-sisternes-epam added panel-review Trigger the apm-review-panel gh-aw workflow and removed status/needs-triage New, awaiting maintainer review. labels May 24, 2026

github-actions Bot removed the panel-review Trigger the apm-review-panel gh-aw workflow label May 24, 2026

danielmeppiel approved these changes May 24, 2026

View reviewed changes

danielmeppiel added this pull request to the merge queue May 24, 2026

github-merge-queue Bot removed this pull request from the merge queue due to failed status checks May 24, 2026

sergio-sisternes-epam mentioned this pull request May 25, 2026

fix: resolve 3 flaky integration tests blocking merge queue #1477

Merged

danielmeppiel added this pull request to the merge queue May 26, 2026

Merged via the queue into microsoft:main with commit c5a63cb May 26, 2026
55 checks passed

danielmeppiel mentioned this pull request May 26, 2026

chore: cut 0.15.0 #1493

Merged

danielmeppiel mentioned this pull request May 27, 2026

docs: backfill apm-usage and consolidate registry guides (v0.14->v0.15 drift sweep) #1511

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(mcp): rewrite skill-shipped MCP command paths for the Claude target#1465

fix(mcp): rewrite skill-shipped MCP command paths for the Claude target#1465
danielmeppiel merged 1 commit into
microsoft:mainfrom
juspay:fix-claude-mcp-skill-command-path

srid commented May 23, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

srid May 23, 2026

Uh oh!

github-actions Bot commented May 24, 2026

Python Architect

CLI Logging Expert

DevX UX Expert

Supply Chain Security Expert

OSS Growth Hacker

Doc Writer

Test Coverage Expert

Auth Expert -- inactive

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

srid commented May 23, 2026

Summary

Repro

Fix

Why this layer

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

srid May 23, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented May 24, 2026

APM Review Panel: ship_with_followups

Panel summary

Top 5 follow-ups

Architecture

Recommendation

Python Architect

CLI Logging Expert

DevX UX Expert

Supply Chain Security Expert

OSS Growth Hacker

Doc Writer

Test Coverage Expert

Auth Expert -- inactive

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

APM Review Panel: `ship_with_followups`