- react-features.md: add interactive-login warning to source maps wizard
reference (P1 — agent would hang on browser auth)
- cocoa SKILL.md: align 'once it finishes' copy with other skills to
point to Verification instead of vague 'continue with configuration'
Co-Authored-By: Claude (Anthropic Claude Sonnet 4)
The Sentry wizard CLI requires interactive browser-based login and
org/project selection prompts that AI agents cannot handle. Updated all
SDK skills that reference the wizard to present it inside a blockquote
telling the user to run the command themselves in their terminal.
If the user skips the wizard, the agent proceeds with full manual setup.
Updated skills:
- sentry-nextjs-sdk
- sentry-svelte-sdk
- sentry-react-native-sdk
- sentry-dotnet-sdk
- sentry-cocoa-sdk
- sentry-sdk-skill-creator (template + philosophy reference)
Co-Authored-By: Claude (Anthropic Claude Sonnet 4)
Extract validate-skill-tree from the warden workflow into a dedicated
.github/workflows/validate-skill-tree.yml. Keeps concerns separate —
warden handles AI code review, skill tree validation handles structural
integrity. Both trigger on the same PR events but run independently.
Use ${ARRAY[@]+"${ARRAY[@]}"} pattern for all array expansions that
could be empty (category arrays, errors array, function parameters).
Bash 3.2 with set -u treats "${EMPTY[@]}" as an unbound variable error,
while bash 4+ handles it gracefully. The +"..." syntax expands to nothing
when the array is unset/empty, avoiding the crash.
Also fix AGENTS.md claiming CI only runs on skills/** changes — the
validate-skill-tree job runs on all PRs (it's fast enough not to need
path filtering).
Adds a parallel CI job that runs `./scripts/build-skill-tree.sh --check`
on every PR. The job checks out the repo with minimal permissions
(contents: read) and fails if SKILL_TREE.md is stale or any skill
frontmatter/breadcrumb/router-table validation errors are found.
Running it as a separate job (rather than a step in the warden review
job) keeps concerns isolated and lets both jobs run in parallel.
Co-Authored-By: Claude Sonnet 4.5 (Anthropic)
Documents the router pattern that keeps agent startup metadata lean
(~300 tokens vs ~1,600+ as the skill library grows). Covers:
- How the three router skills (sentry-sdk-setup, sentry-workflow,
sentry-feature-setup) gate access to hidden child skills via
disable-model-invocation
- Category taxonomy with responsible routers
- Step-by-step checklist for adding a new skill (frontmatter fields,
breadcrumb, router table update, build script)
- Guidance on splitting categories once they exceed ~10 skills
- Reference to scripts/build-skill-tree.sh for local validation and
its --check flag used in CI
This gives contributors (human and AI) a single authoritative source
for skill tree conventions without needing to reverse-engineer the
frontmatter or ask questions.
Co-Authored-By: Claude Sonnet 4.6 (Anthropic)
Adds scripts/build-skill-tree.sh — a bash-only script that scans
all skills/*/SKILL.md files, regenerates SKILL_TREE.md from live
frontmatter data, and validates the entire skill hierarchy.
What the script does:
1. Scan — finds all SKILL.md files via find, parses YAML frontmatter
(name, description, category, parent, role, disable-model-invocation)
using awk. No external dependencies beyond standard Unix tools.
2. Categorize — groups skills by category field, identifies routers
by role: router.
3. Generate — rebuilds SKILL_TREE.md with Quick Navigation table,
per-category skill tables (SDK Setup, Workflow, Feature Setup,
Internal), and short column values extracted from descriptions.
4. Validate — checks every rule without stopping on first failure:
- Non-router/non-internal skills must have category, parent, and
disable-model-invocation: true
- Internal skills must have disable-model-invocation: true
- Every skill's parent must exist and have role: router
- Each skill must appear in its router's SKILL.md
- All breadcrumb .md links must resolve to real files
- Warns on unknown categories
5. Stale check — diffs generated vs existing SKILL_TREE.md; writes
file unless --check is passed (check-only mode for CI).
Uses a temp-directory pattern for per-skill data storage instead of
bash 4 associative arrays, keeping it compatible with macOS bash 3.2.
Also updates SKILL_TREE.md to match the script's generated output
(column values are now extracted from descriptions via sed, ordering
follows find | sort alphabetical).
Co-Authored-By: Claude Sonnet 4.5 (Anthropic)
Creates a flat sitemap at the repo root that gives any agent an instant
overview of the full skill structure. The file contains:
- Quick navigation table mapping user intent to the three router skills
(sentry-sdk-setup, sentry-workflow, sentry-feature-setup)
- Four categorized tables covering all 16 skills: SDK Setup (9 platforms),
Workflow (3 skills), Feature Setup (3 skills), and Internal (1 skill)
- Every skill name and path is a clickable markdown link
- Kept under 1000 tokens as specified
This file serves as the entry point agents use to locate the right skill
before diving into individual SKILL.md files. It will later be generated
automatically by the build script; this is the hand-authored initial version.
Co-Authored-By: Claude Sonnet 4.6 (Anthropic)
Extends the skill-tree navigation pattern (already applied to SDK skills)
to the remaining workflow, feature-setup, and internal skills.
Each file receives three new frontmatter fields:
- category: workflow | feature-setup | internal
- parent: router skill name (omitted for internal/sentry-sdk-skill-creator)
- disable-model-invocation: true
And a breadcrumb blockquote immediately after the closing --- for
contextual navigation:
Workflow skills (parent: sentry-workflow):
sentry-code-review, sentry-pr-code-review, sentry-fix-issues
Feature-setup skills (parent: sentry-feature-setup):
sentry-setup-ai-monitoring, sentry-otel-exporter-setup, sentry-create-alert
Internal skills (no parent router):
sentry-sdk-skill-creator → breadcrumb links to SKILL_TREE.md only
No existing skill content was modified — pure additive changes.
Co-Authored-By: Claude Sonnet 4.5 (Anthropic)
Add category, parent, and disable-model-invocation fields to the YAML
frontmatter of all 9 SDK skill files, and insert a breadcrumb navigation
line as the first body element after the closing ---.
Each skill now declares:
category: sdk-setup
parent: sentry-sdk-setup
disable-model-invocation: true
And opens with a breadcrumb blockquote:
> [All Skills](../../SKILL_TREE.md) > [SDK Setup](../sentry-sdk-setup/SKILL.md) > <Platform> SDK
This establishes the skill tree hierarchy for the sdk-setup category,
enabling routers and tooling to discover parent/child relationships and
preventing these implementation skills from being invoked directly by
the model without going through the sentry-sdk-setup router.
No existing skill content was modified — only additive changes.
Co-Authored-By: Claude Sonnet 4.5 (Anthropic)
Introduce three new router-role skills that act as entry points for common
Sentry tasks, routing agents to the appropriate specialized skill based on
user intent and detected platform context.
- sentry-sdk-setup: detects project platform from files (package.json,
go.mod, Gemfile, *.xcodeproj, etc.) and maps to one of 9 SDK skills
(Cocoa, .NET, Go, Next.js, Python, React Native, React, Ruby, Svelte).
Includes priority rules for Next.js over React and React Native over React,
with a fallback to docs.sentry.io/platforms/ for unmatched platforms.
- sentry-workflow: routes debugging and code review tasks to sentry-fix-issues
(MCP-based production debugging), sentry-code-review (sentry[bot] PR
comments), or sentry-pr-code-review (Seer Bug Prediction reviews).
- sentry-feature-setup: routes feature configuration requests to
sentry-setup-ai-monitoring (LLM/AI SDK instrumentation),
sentry-otel-exporter-setup (OTel Collector pipelines), or
sentry-create-alert (workflow engine alerts and notifications).
All three skills use role: router in frontmatter and include breadcrumb
navigation back to SKILL_TREE.md.
Co-Authored-By: Claude Sonnet 4.5 (Anthropic AI assistant)
Restructure the README around what users can actually do with the
plugin rather than listing technical components. Opens with the core
value prop — your AI assistant knows code, this teaches it Sentry — then
shows real examples for each use case before diving into skill tables.
Drops the plugin structure tree (internal detail) and streamlines
installation instructions. Adds a contributing section pointing to the
agent skills spec and the sdk-skill-creator skill.
Co-Authored-By: Claude (Anthropic)
Update documentation to reflect the full skill catalog ported from
sentry-agent-skills. Skills are now organized into four categories:
- SDK bundles (9): full platform setup wizards with reference docs
- Setup skills (2): AI monitoring, OTel exporter
- Workflow skills (4): code review, PR review, fix issues, create alerts
- Authoring skills (1): SDK skill creator
Also adds scripts/generate-skill-list.sh from sentry-agent-skills for
generating skill lists from SKILL.md frontmatter.
Co-Authored-By: Claude (Anthropic)
Remove 5 legacy skills that are being replaced by comprehensive SDK
skill bundles from sentry-agent-skills:
- sentry-ios-swift-setup: replaced by sentry-cocoa-sdk
- sentry-setup-ai-monitoring: replaced by updated version
- sentry-setup-logging: covered by per-SDK bundles (Python, Ruby, etc.)
- sentry-setup-metrics: covered by per-SDK bundles
- sentry-setup-tracing: covered by per-SDK bundles
The sentry-code-review skill is retained as-is.
Co-Authored-By: Claude (Anthropic)
Merges `sentry-for-claude` and `sentry-for-cursor` into a single repo that serves both platforms. Both repos contained the same Sentry plugin (MCP server, skills, commands) packaged for different platforms — maintaining two repos meant duplicating every change.
Skills and commands now live at the repo root where both Claude Code and Cursor auto-discover them. Each platform gets its own metadata directory (`.claude-plugin/`, `.cursor-plugin/`) with separate `plugin.json` and `marketplace.json` files — no symlinks between platform-specific configs. MCP configs are also separate files (`.mcp.json` for Claude Code, `mcp.json` for Cursor) since the platforms use different file conventions.
`.agents/` is preserved as symlinks to `commands/` and `skills/` for backwards compatibility.
Plugin manifests validated against the official schemas from [cursor/plugins](https://github.com/cursor/plugins/tree/main/schemas) and [anthropics/claude-code](https://github.com/anthropics/claude-code).
After merge, rename repo from `sentry-for-claude` to `sentry-for-ai` and archive `sentry-for-cursor` with a redirect notice.
---------
Co-authored-by: Claude <noreply@anthropic.com>
The workflow triggers on pull_request types [opened, synchronize, reopened]
but the skill trigger only matched opened and synchronize, causing reopened
PRs to run the workflow without triggering the skill-scanner scan.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
These capabilities are being rebuilt as part of the getsentry/sentry-mcp
server. Also removes the now-redundant MCP-SETUP.md since MCP is
auto-configured via .mcp.json and plugin.json.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove redundant code examples, consolidate feature sections, and
restructure to follow the same pattern as existing setup skills.
Reduced from 735 to 391 lines while preserving all essential content.
Co-Authored-By: Claude (claude-opus-4-6)
Add sentry-ios-swift-setup skill for configuring Sentry in iOS/Swift
apps including error monitoring, tracing, session replay, logging,
and profiling. Update README and AGENTS.md with complete skills listing.
Co-Authored-By: Claude (claude-opus-4-6)
Daniel Griesser