feat: v2.0 — orchestration patterns, reference guides, cross-agent support#23
Merged
feat: v2.0 — orchestration patterns, reference guides, cross-agent support#23
Conversation
…ross-agent support New reference guides: - docs/settings-guide.md: Complete settings hierarchy, permissions, sandbox, env vars - docs/cli-cheatsheet.md: Every CLI flag, keyboard shortcut, slash command - docs/orchestration-patterns.md: Command > Agent > Skill architecture with examples - docs/context-loading.md: CLAUDE.md monorepo loading, agent memory, skills discovery - docs/cross-agent-workflows.md: Claude Code + Cursor + Codex config mapping - docs/new-features.md: Voice mode, agent teams, checkpointing, new hook events - docs/daily-habits.md: Session habits, debugging tips, terminal setup New agents: - agents/orchestrator.md: Multi-phase dev (Research > Plan > Implement) - agents/debugger.md: Systematic hypothesis-driven bug investigation New commands: - commands/develop.md: Multi-phase feature build with validation gates - commands/doctor.md: Health check for pro-workflow setup New skills: - skills/context-optimizer: Token management and context budget planning - skills/orchestrate: Command > Agent > Skill wiring guide Enhanced hooks (18 events, up from 11): - SubagentStart/SubagentStop: Agent lifecycle observability - TaskCompleted: Quality gate on task completion - PermissionRequest: Flag dangerous operations - PostToolUseFailure: Track tool failures - TeammateIdle: Detect blockers in agent teams New configs: - settings.example.json: Production-ready settings template - Updated mcp-config.example.json: Curated recommendations (context7, playwright, github) Updated: - README.md: Complete rewrite for v2.0 - SKILL.md: Added orchestration pattern, daily habits, reference docs - references/claude-code-resources.md: Community workflows and learning path - plugin.json: v2.0, added new agents - package.json: v2.0, added docs to files - marketplace.json: Updated description and tags
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (4)
📝 WalkthroughWalkthroughBumps project to v2.0.0, updates plugin manifests and package metadata, adds two agent docs, many new reference guides and skills, expands hooks and MCP example, and adds a settings example — documentation and manifest updates only; no functional source code changes. Changes
Sequence Diagram(s)(omitted) Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…l demos, mermaid charts Visual assets: - assets/banner.svg: Dark theme hero banner with stats badges and terminal preview - assets/architecture.svg: Full architecture diagram showing Commands > Agents > Skills > Hooks layers - assets/workflow-flow.svg: /develop multi-phase flow with validation gates and hook events - assets/terminal-demo.svg: Catppuccin-themed terminal showing /develop in action - assets/self-correction-demo.svg: Terminal demo of the self-correction loop README enhancements: - Hero banner with styled badges (stars, npm, license, SkillKit) - Centered header with feature summary - Self-correction and /develop demo screenshots inline - Architecture diagram section - Mermaid hook lifecycle diagram (GitHub-native rendering) - Mermaid cross-agent support graph - Quick start section at bottom - Styled footer with centered badges and links
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 16
🧹 Nitpick comments (7)
references/claude-code-resources.md (1)
47-47: Normalize article source formatting to full URLs.Line 47 uses
cursor.com/blog/agent-best-practiceswhile neighboring entries use named sources or clearer URL forms. Consider using a full link for consistency and clickability.Suggested fix
-| Agent Best Practices | cursor.com/blog/agent-best-practices | +| Agent Best Practices | https://cursor.com/blog/agent-best-practices |🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@references/claude-code-resources.md` at line 47, Replace the partial link entry "Agent Best Practices | cursor.com/blog/agent-best-practices" with a normalized, full URL form (e.g., "Agent Best Practices | https://cursor.com/blog/agent-best-practices") to match the surrounding entries' formatting and ensure clickability; update the exact table row text so the source appears as a full URL.skills/orchestrate/SKILL.md (1)
12-26: Add language specifier to the workflow diagram.The ASCII diagram should have a language identifier (e.g.,
text) for consistent rendering.📝 Suggested fix
-``` +```text /develop <feature> │🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@skills/orchestrate/SKILL.md` around lines 12 - 26, The ASCII workflow block in SKILL.md (the triple-backtick fenced diagram starting with "/develop <feature>") lacks a language specifier; update the fenced code block to use a text language label (e.g., change ``` to ```text at the opening fence and keep the closing ```), so the diagram consistently renders as plain text..claude-plugin/plugin.json (1)
4-4: Consider updating the description to reflect v2.0 capabilities.The description still references the v1.x feature set. Consider aligning it with the updated
package.jsondescription that includes "orchestration patterns" and "cross-agent support".🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.claude-plugin/plugin.json at line 4, Update the "description" value in .claude-plugin/plugin.json (the JSON key "description") to reflect v2.0 capabilities by replacing the v1.x phrasing with the updated package.json language — include mentions of "orchestration patterns" and "cross-agent support" and refer to persistent storage, searchable learnings, and new v2 features so the description aligns with package.json and accurately brands this as v2.0.README.md (2)
211-251: Add language specifiers to fenced code blocks.The structure tree and database examples use unspecified fenced code blocks. Adding language identifiers improves rendering and accessibility.
📝 Suggested fix for lines 211, 257, 262
## Structure -``` +```text pro-workflow/ ├── .claude-plugin/ # Claude Code plugin-``` +```text ~/.pro-workflow/ └── data.db # SQLite database with learnings and sessions-``` +```bash /search testing # Find all testing-related learnings🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` around lines 211 - 251, Update the three unspecified fenced code blocks in README.md: add a language specifier "text" to the pro-workflow directory tree block (the block starting with "pro-workflow/"), add "text" to the ~/.pro-workflow/ SQLite example block (the block with "data.db"), and add "bash" to the command example block (the block starting "/search testing"); locate these blocks by those exact snippet headers and replace the opening ``` with ```text or ```bash accordingly so renderers and accessibility tools get the correct language metadata.
190-194: Capitalize "GitHub" for consistency.Per brand guidelines, "GitHub" should be capitalized.
📝 Suggested fix
-- **github** — PRs, issues, code search +- **GitHub** — PRs, issues, code search🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` around lines 190 - 194, Update the list item that currently reads "**github** — PRs, issues, code search" to use the proper brand capitalization "GitHub"; locate the Markdown bullet with the symbol **github** and replace it with **GitHub** so the README entry becomes "**GitHub** — PRs, issues, code search".agents/orchestrator.md (1)
40-61: Add language specifier to the plan template code block.The fenced code block should have a language identifier for better rendering.
📝 Suggested fix
-``` +```text PLAN: [Feature Name]🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@agents/orchestrator.md` around lines 40 - 61, Update the fenced code block that starts with "PLAN: [Feature Name]" to include a language specifier (e.g., add "text" after the opening backticks) so the block renders with the correct language; modify the block in agents/orchestrator.md where the PLAN template is defined (the fenced block containing "Goal: [one sentence]" and "Files to modify:") to use ```text instead of ``` to ensure proper rendering.docs/cross-agent-workflows.md (1)
24-24: Avoid hardcoded pricing in evergreen docs.
~$40/monthis likely to drift; consider “pricing varies by plan/region” (or timestamp it).🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/cross-agent-workflows.md` at line 24, Replace the hardcoded sentence "Cost: ~$40/month combined." with a neutral, evergreen phrasing (e.g., "Pricing varies by plan and region" or "Estimated cost: varies by plan/region; check provider pricing (date)"), or add a timestamped note, so the docs avoid stale fixed pricing and direct readers to current provider pricing.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@agents/debugger.md`:
- Around line 25-68: Add explicit language identifiers to the three fenced code
blocks in agents/debugger.md to satisfy MD040: change the fences for the
"Hypothesis" block, the "ROOT CAUSE" block, and the "FIX" block to start with
```text (rather than ```); update the three blocks that begin at the Hypothesis
section, the Root Cause section, and the Fix Proposal section so each fenced
block declares "text" as its language. This is a minimal markdown-only change
and ensures linting passes.
In `@commands/develop.md`:
- Around line 35-54: The fenced code block containing the "PLAN: [Feature Name]"
template should include a language tag to satisfy markdownlint; update the
opening fence from ``` to ```text (or ```md) for the fenced plan template block
in commands/develop.md so the linter recognizes the block as plain text/markdown
and the lint warning at the template block (around the "PLAN: [Feature Name]"
section) is resolved.
In `@commands/doctor.md`:
- Around line 29-31: The unlabeled fenced code blocks that contain "/context"
(and the other block covering lines 77-85) must include a language identifier to
satisfy MD040; update those triple-backtick fences to use an appropriate label
such as ```bash for shell/command examples or ```text for plain output so both
the `/context` block and the block spanning 77-85 are labeled accordingly.
In `@docs/context-loading.md`:
- Around line 11-22: Several fenced code blocks in docs/context-loading.md
(notably the ASCII directory tree and the other example blocks shown around the
ASCII tree and later examples) are missing language identifiers; update each
triple-backtick fence to include an appropriate language tag (e.g., ```text for
the ASCII tree, ```bash for shell snippets, or ```markdown for markdown
examples) so they satisfy MD040 and render correctly. Locate the examples that
correspond to the blocks shown in the diff (the directory tree block and the
other sample blocks referenced in the review) and add the correct fence language
to each opening ``` fence.
In `@docs/cross-agent-workflows.md`:
- Around line 18-22: The fenced code block shown (the triple-backtick block
starting at the unlabeled diagram lines) needs a language label to satisfy
markdown linting—edit the opening fence from ``` to ```text so the block is a
labeled "text" code fence; locate the diagram block containing "Cursor for daily
coding (tab completions, inline edits)" and update the opening backticks
accordingly.
In `@docs/daily-habits.md`:
- Around line 101-110: The unlabeled fenced block that starts with "Primary
editor (Cursor/VS Code)" should be marked with the language specifier text to
satisfy MD040 and ensure deterministic formatting—update the opening
triple-backtick for that block to include text (i.e., label the code fence with
"text") and leave the existing content unchanged.
In `@docs/new-features.md`:
- Around line 11-13: The fenced code block that contains "/voice # Toggle
voice mode on/off" is unlabeled and triggers MD040; update that fenced block
(the triple-backtick fence surrounding the "/voice" snippet in
docs/new-features.md) to include the language label "bash" (i.e., change ``` to
```bash) so the block is properly labeled for Markdown linting.
- Around line 121-145: The docs list events and hook types that aren't present
in the runtime config; either update the documentation or the configuration so
they match. Either remove `Setup`, `WorktreeCreate`, `WorktreeRemove` and the
`prompt`/`agent` types from docs/new-features.md, or add corresponding entries
to hooks/hooks.json (create hook definitions for `Setup`, `WorktreeCreate`,
`WorktreeRemove` and add hook type entries for `prompt` and `agent`), ensuring
the hook schema and any validation logic accept those types and event names
(reference the event names `SubagentStart`, `SubagentStop`, `Setup`,
`PermissionRequest`, `TeammateIdle`, `TaskCompleted`, `ConfigChange`,
`WorktreeCreate`, `WorktreeRemove`, `PostToolUseFailure` and the types
`command`, `prompt`, `agent` to locate where to change docs or
hooks/hooks.json).
In `@docs/orchestration-patterns.md`:
- Around line 7-130: The three plain-text fenced blocks in
docs/orchestration-patterns.md (the ASCII diagram starting with "Command (entry
point, user-facing)", the repository tree block beginning with ".claude/", and
the flow block starting with "/develop \"add webhook support\"") are missing a
language tag; update each opening triple-backtick to include the language tag
`text` (e.g., ```text) so markdownlint recognizes them as plain-text code
blocks.
In `@docs/settings-guide.md`:
- Around line 7-14: The fenced code block in docs/settings-guide.md (the block
showing the numbered settings list) is unlabeled; update the opening fence to
include a language label (use "text") so the block becomes a labeled code fence
(e.g., change ``` to ```text) to satisfy MD040 and ensure proper
linting/rendering.
In `@hooks/hooks.json`:
- Around line 198-208: The PermissionRequest hook currently embeds a hardcoded
dangerous list and uses JSON.stringify(i).includes(d) which causes divergence
from settings.example.json and false positives; update the command in the
PermissionRequest hook to (1) align the dangerous patterns with the official
rules (e.g., match "git push *", "git reset *", "npm publish *" rather than
crude substrings like "publish" or "push --force") and (2) replace the loose
JSON.stringify substring check with explicit parsing of the incoming object (use
i.tool and i.args or i.command fields) and safe pattern matching (regular
expressions or exact token comparisons) so only intended commands match;
reference the PermissionRequest hook entry and the embedded command string when
making the changes.
In `@references/claude-code-resources.md`:
- Around line 73-75: Add a language identifier to the fenced code block that
contains "[LEARN] Claude-Code: <the practice>" so markdownlint MD040 is
satisfied and syntax highlighting is preserved; update the opening
triple-backtick for that block (the block around the string "[LEARN]
Claude-Code: <the practice>") to include a language token such as "text" (e.g.,
```text).
In `@settings.example.json`:
- Line 55: The example config currently sets the "enableAllProjectMcpServers"
flag to true which grants broad MCP trust by default; change the default value
to false in settings.example.json and add a brief comment above or near the
"enableAllProjectMcpServers" key indicating that enabling it is a privileged,
opt-in action for production deployments to encourage least-privilege usage.
- Line 49: The permissions array contains an undocumented entry "Skill(*)" that
should either be removed from settings.example.json or documented in
docs/settings-guide.md; update settings.example.json to replace "Skill(*)" with
a documented example like "Agent(*)" if you don't intend to support skill
wildcards, or add a new "Skill" wildcard rule under the Permission
Rules/wildcard table in docs/settings-guide.md explaining the behavior and scope
of "Skill(*)" (mirror the format and examples used for "Task(*)" and
"Agent(*)").
In `@skills/context-optimizer/SKILL.md`:
- Around line 42-45: The fenced code block in SKILL.md containing the lines
"/mcp # List active servers" and "# Disable what you're not using"
lacks a language tag and triggers MD040; update that fenced block (the
triple-backtick block surrounding those lines) to include a language specifier
such as bash (e.g., change ``` to ```bash) so markdownlint recognizes it as a
shell snippet.
In `@skills/pro-workflow/SKILL.md`:
- Around line 445-449: The unlabeled fenced code block that begins with the
triple backticks before the line "Command (user-facing entry point)" needs a
language label to satisfy MD040; edit the opening fence from ``` to ```text so
the block reads as a text code fence and leave the block content unchanged.
---
Nitpick comments:
In @.claude-plugin/plugin.json:
- Line 4: Update the "description" value in .claude-plugin/plugin.json (the JSON
key "description") to reflect v2.0 capabilities by replacing the v1.x phrasing
with the updated package.json language — include mentions of "orchestration
patterns" and "cross-agent support" and refer to persistent storage, searchable
learnings, and new v2 features so the description aligns with package.json and
accurately brands this as v2.0.
In `@agents/orchestrator.md`:
- Around line 40-61: Update the fenced code block that starts with "PLAN:
[Feature Name]" to include a language specifier (e.g., add "text" after the
opening backticks) so the block renders with the correct language; modify the
block in agents/orchestrator.md where the PLAN template is defined (the fenced
block containing "Goal: [one sentence]" and "Files to modify:") to use ```text
instead of ``` to ensure proper rendering.
In `@docs/cross-agent-workflows.md`:
- Line 24: Replace the hardcoded sentence "Cost: ~$40/month combined." with a
neutral, evergreen phrasing (e.g., "Pricing varies by plan and region" or
"Estimated cost: varies by plan/region; check provider pricing (date)"), or add
a timestamped note, so the docs avoid stale fixed pricing and direct readers to
current provider pricing.
In `@README.md`:
- Around line 211-251: Update the three unspecified fenced code blocks in
README.md: add a language specifier "text" to the pro-workflow directory tree
block (the block starting with "pro-workflow/"), add "text" to the
~/.pro-workflow/ SQLite example block (the block with "data.db"), and add "bash"
to the command example block (the block starting "/search testing"); locate
these blocks by those exact snippet headers and replace the opening ``` with
```text or ```bash accordingly so renderers and accessibility tools get the
correct language metadata.
- Around line 190-194: Update the list item that currently reads "**github** —
PRs, issues, code search" to use the proper brand capitalization "GitHub";
locate the Markdown bullet with the symbol **github** and replace it with
**GitHub** so the README entry becomes "**GitHub** — PRs, issues, code search".
In `@references/claude-code-resources.md`:
- Line 47: Replace the partial link entry "Agent Best Practices |
cursor.com/blog/agent-best-practices" with a normalized, full URL form (e.g.,
"Agent Best Practices | https://cursor.com/blog/agent-best-practices") to match
the surrounding entries' formatting and ensure clickability; update the exact
table row text so the source appears as a full URL.
In `@skills/orchestrate/SKILL.md`:
- Around line 12-26: The ASCII workflow block in SKILL.md (the triple-backtick
fenced diagram starting with "/develop <feature>") lacks a language specifier;
update the fenced code block to use a text language label (e.g., change ``` to
```text at the opening fence and keep the closing ```), so the diagram
consistently renders as plain text.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 61e57e72-638b-4f2a-acf8-a28499c849ab
📒 Files selected for processing (22)
.claude-plugin/marketplace.json.claude-plugin/plugin.jsonREADME.mdagents/debugger.mdagents/orchestrator.mdcommands/develop.mdcommands/doctor.mddocs/cli-cheatsheet.mddocs/context-loading.mddocs/cross-agent-workflows.mddocs/daily-habits.mddocs/new-features.mddocs/orchestration-patterns.mddocs/settings-guide.mdhooks/hooks.jsonmcp-config.example.jsonpackage.jsonreferences/claude-code-resources.mdsettings.example.jsonskills/context-optimizer/SKILL.mdskills/orchestrate/SKILL.mdskills/pro-workflow/SKILL.md
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@README.md`:
- Line 265: Replace the lowercase bold token "**github**" in the README with the
official product casing "**GitHub**" so the line reads "**GitHub** — PRs,
issues, code search"; update the exact bolded text wherever "**github**" appears
to maintain consistent product casing.
- Around line 304-344: Update the fenced code blocks in the README by adding
language identifiers: mark the directory tree blocks (the "pro-workflow/" tree
and the "~/.pro-workflow/" tree) with ```text and mark the command examples (the
`/search testing` and `/search "file paths"` block) with ```bash so the three
shown fenced blocks are lint-compliant (also apply the same changes to the other
similar blocks referenced in the comment).
- Around line 14-15: The README states "18 hook events" but hooks/hooks.json
only defines 15 unique hook keys; update the README to reflect the actual count
(change "18 hook events" to "15 hook events") wherever it appears (the other
occurrences flagged in the comment) after verifying hooks/hooks.json for the
definitive list, or alternatively add the missing hook definitions to
hooks/hooks.json if you intend to keep "18"; update all README mentions
(including the other flagged sections) to be consistent with hooks/hooks.json
and ensure the phrasing matches the verified hook count.
- Around line 174-188: The Commands table lists namespaced commands like
/pro-workflow:wrap-up while Quick Start shows short forms like /develop,
/doctor, /wrap-up; update the README so a single canonical invocation format is
used (either fully namespaced /pro-workflow:<command> or the short forms
/develop, /doctor, /wrap-up) and make this consistent across the Commands table
and Quick Start; if both forms are supported, explicitly document each alias
(e.g., “/pro-workflow:wrap-up (alias: /wrap-up)”) near the table and the Quick
Start examples and ensure all entries (including /pro-workflow:learn-rule,
/pro-workflow:commit, /pro-workflow:develop, /pro-workflow:doctor,
/pro-workflow:insights, /pro-workflow:replay, /pro-workflow:handoff,
/pro-workflow:search, /pro-workflow:list) reflect the chosen canonical format.
- Around line 30-31: Update the README's `/develop` workflow flow text to match
the command behavior by adding the missing "Review & Commit" phase so it reads
"Research > Plan > Implement > Review & Commit" (and update the second
occurrence of the same flow elsewhere in the README); ensure the wording matches
the phase name used in commands/develop.md and any other mentions of the
`/develop` command.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 6d98ff45-d772-4936-9ba5-17d756346d31
⛔ Files ignored due to path filters (5)
assets/architecture.svgis excluded by!**/*.svgassets/banner.svgis excluded by!**/*.svgassets/self-correction-demo.svgis excluded by!**/*.svgassets/terminal-demo.svgis excluded by!**/*.svgassets/workflow-flow.svgis excluded by!**/*.svg
📒 Files selected for processing (1)
README.md
- Add language identifiers to ~25 fenced code blocks (MD040) across agents, commands, docs, skills, references, and README - Fix PermissionRequest hook: use proper field parsing instead of JSON.stringify substring matching for security - Change enableAllProjectMcpServers to false in settings.example.json and settings-guide.md (safer default) - Remove unnecessary Skill(*) permission rule from settings.example.json - Update plugin.json description to reflect v2.0 scope - Add "In hooks.json" column to new-features.md hook events table to clarify which events are configured vs available - Replace hardcoded pricing in cross-agent-workflows.md - Add full URL to cursor.com blog link in resources
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
skills/pro-workflow/SKILL.md (1)
154-154:⚠️ Potential issue | 🟡 MinorUnify planning/orchestration thresholds (
>3vs>5files).This doc currently defines two different cutoffs for escalation, which can produce inconsistent behavior. Pick one threshold or explain the distinction explicitly.
Also applies to: 421-422
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@skills/pro-workflow/SKILL.md` at line 154, The document currently uses two different escalation thresholds (">3" in the "Plan mode when: >3 files, architecture decisions, multiple approaches." line and a different cutoff at lines 421-422); standardize by choosing a single numeric threshold (e.g., use ">3 files" everywhere) or explicitly document the distinction between the two thresholds (e.g., "Use >3 files for minor multi-file changes; use >5 files for large refactors/architectural decisions") and update both the "Plan mode" line and the content at lines 421-422 to reflect the chosen rule and wording consistently.
♻️ Duplicate comments (5)
docs/new-features.md (1)
121-122:⚠️ Potential issue | 🟡 MinorClarify platform capability vs template default hook count.
This sentence is ambiguous against current config (
hooks/hooks.jsondefines 15 groups). Reword so it clearly states Claude Code capability vs what this template enables by default.✍️ Suggested wording
-Claude Code has expanded from 8 to 18+ hook event types. +Claude Code supports 18+ hook event types; this template enables 15 by default.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/new-features.md` around lines 121 - 122, The sentence in docs/new-features.md ambiguously mixes platform capability with the template's default hook count; update the sentence to explicitly state that Claude Code supports 18+ hook event types while this template enables 15 groups by default (as defined in hooks/hooks.json), e.g., split into two clear clauses: one declaring platform capability and one declaring the template/default configuration and referencing hooks/hooks.json for the 15 groups.hooks/hooks.json (1)
204-204:⚠️ Potential issue | 🟡 MinorDangerous-operation matching is still brittle and can drift from rule intent.
The
includes-based check can produce false positives/negatives and still doesn’t model command intent precisely (git push *,git reset *, etc.). Use command-focused regex checks instead of broad substrings.🛡️ Suggested matcher refinement
-const dangerous=['rm -rf','docker','deploy','npm publish','push --force','reset --hard']; -const isDangerous=dangerous.some(p=>tool.includes(p)||cmd.includes(p)); +const dangerous=[ + /\brm\s+-rf\b/, + /\bgit\s+push\b/, + /\bgit\s+reset\b/, + /\bdocker\b/, + /\bnpm\s+publish\b/, + /\bdeploy\b/ +]; +const isDangerous=dangerous.some((re)=>re.test(cmd));🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@hooks/hooks.json` at line 204, Replace the brittle substring check that builds the "dangerous" array and computes isDangerous with regex-based, command-focused matchers: instead of dangerous.some(p=>tool.includes(p)||cmd.includes(p)), define a list of anchored regex patterns for tool and cmd (e.g., patterns that match whole commands like /^git\s+push\b/, /^git\s+reset\b/, /^rm\s+-rf\b/, /^docker\b/, /^npm\s+publish\b/, /--force\b/ etc.) and test tool and command separately against those regexes; update the inline node script's variables (the existing dangerous array and isDangerous computation) to iterate regexes and use RegExp.test on tool and on cmd to decide danger so intent is matched more precisely and false positives from simple substrings are avoided.README.md (3)
30-31:⚠️ Potential issue | 🟡 Minor
/developphase flow omits the final review phase.The README still documents
Research > Plan > Implement, but the command includesReview & Commit. Please update all occurrences to the full phase chain.Also applies to: 87-87, 181-181
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` around lines 30 - 31, Update the README occurrences that describe the /develop command flow to include the missing final phase so the phase chain reads "Research > Plan > Implement > Review & Commit"; search for all mentions of the /develop command and replace the short form "Research > Plan > Implement" (notably the current table row for <code>/develop</code> and the other referenced occurrences) with the full "Research > Plan > Implement > Review & Commit" string so documentation matches the command behavior.
176-188:⚠️ Potential issue | 🟠 MajorDocument one canonical command format (or explicitly list aliases).
README mixes namespaced commands (
/pro-workflow:*) and short forms (/develop,/doctor,/wrap-up) without stating alias behavior. Standardize this to avoid invalid command usage.Also applies to: 385-387
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` around lines 176 - 188, The README currently mixes namespaced commands like /pro-workflow:develop and short forms like /develop without explaining alias behavior; update the command table to either standardize on one canonical format (e.g., always use /pro-workflow:*) or explicitly list every allowed alias for each action (for example show both /pro-workflow:develop and /develop for the develop command), and add a single-sentence note above the table clarifying whether short forms are supported and how they map to namespaced commands; ensure entries for /pro-workflow:wrap-up, /pro-workflow:learn-rule, /pro-workflow:commit, /pro-workflow:develop, /pro-workflow:doctor, /pro-workflow:insights, /pro-workflow:replay, /pro-workflow:handoff, /pro-workflow:search, and /pro-workflow:list are all updated consistently.
265-265:⚠️ Potential issue | 🟡 MinorUse official product casing:
GitHub.Replace
**github**with**GitHub**.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` at line 265, Replace the lowercase project name formatting in the README: find the markdown token "**github** — PRs, issues, code search" (the bolded "github") and change it to the official product casing "**GitHub**" so the README uses the correct trademark capitalization.
🧹 Nitpick comments (1)
agents/orchestrator.md (1)
3-3: Align the proactive trigger threshold with project planning guidance.This uses
>5 files, while project guidance uses>3 filesfor plan-first execution. Keeping one threshold avoids inconsistent orchestration behavior.♻️ Suggested edit
-description: Multi-phase development agent. Research > Plan > Implement with validation gates. Use PROACTIVELY when building features that touch >5 files or require architecture decisions. +description: Multi-phase development agent. Research > Plan > Implement with validation gates. Use PROACTIVELY when building features that touch >3 files or require architecture decisions.Based on learnings: For tasks touching more than 3 files or involving architectural decisions, enter plan mode first, present an implementation plan, and wait for explicit 'proceed' or 'go ahead' before executing step by step.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@agents/orchestrator.md` at line 3, The orchestration description uses a threshold of ">5 files" for PROACTIVELY triggering plan-first mode but project guidance requires ">3 files"; update the description string in the orchestrator (the line containing "Multi-phase development agent... Use PROACTIVELY when building features that touch >5 files") to read ">3 files" and augment the PROACTIVELY trigger wording to state "For tasks touching more than 3 files or involving architectural decisions, enter plan mode first, present an implementation plan, and wait for explicit 'proceed' or 'go ahead' before executing" so the PROACTIVELY behavior and threshold (refer to the description text and the PROACTIVELY trigger wording) are consistent.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.claude-plugin/plugin.json:
- Line 4: The plugin metadata "description" field in .claude-plugin/plugin.json
incorrectly states "18 hooks" while the actual configured hook groups are 15;
update the description string to reflect the shipped configuration (e.g., change
"18 hooks" to "15 hooks") so the text matches hooks/hooks.json and any related
documentation entries; edit the "description" value in
.claude-plugin/plugin.json and run a quick grep or open hooks/hooks.json to
confirm the hook group count before committing.
In `@commands/develop.md`:
- Around line 2-3: Update the frontmatter 'description' string to match the
implemented workflow by listing all four phases (Research, Plan, Implement,
Review & Commit) instead of only three; edit the 'description' value in the file
so it explicitly includes "Research > Plan > Implement > Review & Commit" (keep
'argument-hint' unchanged).
In `@docs/context-loading.md`:
- Around line 131-132: The doc incorrectly implies
SLASH_COMMAND_TOOL_CHAR_BUDGET is a pro-workflow project config key; clarify
that SLASH_COMMAND_TOOL_CHAR_BUDGET is an external runtime environment variable
(set in your shell or the Claude/runtime environment), not a setting in
pro-workflow's config.json. Update the sentence around
SLASH_COMMAND_TOOL_CHAR_BUDGET to explicitly state it must be exported/defined
in your environment (e.g., shell or runtime env) and not placed in project
config files so readers won’t search for it in pro-workflow config.
In `@skills/pro-workflow/SKILL.md`:
- Line 484: Update the product casing in the markdown list item that currently
contains the inline code token `github` to use the official name `GitHub`
(replace the token `` `github` `` with `` `GitHub` ``) so the entry reads
"`GitHub` — PRs, issues, code search".
---
Outside diff comments:
In `@skills/pro-workflow/SKILL.md`:
- Line 154: The document currently uses two different escalation thresholds
(">3" in the "Plan mode when: >3 files, architecture decisions, multiple
approaches." line and a different cutoff at lines 421-422); standardize by
choosing a single numeric threshold (e.g., use ">3 files" everywhere) or
explicitly document the distinction between the two thresholds (e.g., "Use >3
files for minor multi-file changes; use >5 files for large
refactors/architectural decisions") and update both the "Plan mode" line and the
content at lines 421-422 to reflect the chosen rule and wording consistently.
---
Duplicate comments:
In `@docs/new-features.md`:
- Around line 121-122: The sentence in docs/new-features.md ambiguously mixes
platform capability with the template's default hook count; update the sentence
to explicitly state that Claude Code supports 18+ hook event types while this
template enables 15 groups by default (as defined in hooks/hooks.json), e.g.,
split into two clear clauses: one declaring platform capability and one
declaring the template/default configuration and referencing hooks/hooks.json
for the 15 groups.
In `@hooks/hooks.json`:
- Line 204: Replace the brittle substring check that builds the "dangerous"
array and computes isDangerous with regex-based, command-focused matchers:
instead of dangerous.some(p=>tool.includes(p)||cmd.includes(p)), define a list
of anchored regex patterns for tool and cmd (e.g., patterns that match whole
commands like /^git\s+push\b/, /^git\s+reset\b/, /^rm\s+-rf\b/, /^docker\b/,
/^npm\s+publish\b/, /--force\b/ etc.) and test tool and command separately
against those regexes; update the inline node script's variables (the existing
dangerous array and isDangerous computation) to iterate regexes and use
RegExp.test on tool and on cmd to decide danger so intent is matched more
precisely and false positives from simple substrings are avoided.
In `@README.md`:
- Around line 30-31: Update the README occurrences that describe the /develop
command flow to include the missing final phase so the phase chain reads
"Research > Plan > Implement > Review & Commit"; search for all mentions of the
/develop command and replace the short form "Research > Plan > Implement"
(notably the current table row for <code>/develop</code> and the other
referenced occurrences) with the full "Research > Plan > Implement > Review &
Commit" string so documentation matches the command behavior.
- Around line 176-188: The README currently mixes namespaced commands like
/pro-workflow:develop and short forms like /develop without explaining alias
behavior; update the command table to either standardize on one canonical format
(e.g., always use /pro-workflow:*) or explicitly list every allowed alias for
each action (for example show both /pro-workflow:develop and /develop for the
develop command), and add a single-sentence note above the table clarifying
whether short forms are supported and how they map to namespaced commands;
ensure entries for /pro-workflow:wrap-up, /pro-workflow:learn-rule,
/pro-workflow:commit, /pro-workflow:develop, /pro-workflow:doctor,
/pro-workflow:insights, /pro-workflow:replay, /pro-workflow:handoff,
/pro-workflow:search, and /pro-workflow:list are all updated consistently.
- Line 265: Replace the lowercase project name formatting in the README: find
the markdown token "**github** — PRs, issues, code search" (the bolded "github")
and change it to the official product casing "**GitHub**" so the README uses the
correct trademark capitalization.
---
Nitpick comments:
In `@agents/orchestrator.md`:
- Line 3: The orchestration description uses a threshold of ">5 files" for
PROACTIVELY triggering plan-first mode but project guidance requires ">3 files";
update the description string in the orchestrator (the line containing
"Multi-phase development agent... Use PROACTIVELY when building features that
touch >5 files") to read ">3 files" and augment the PROACTIVELY trigger wording
to state "For tasks touching more than 3 files or involving architectural
decisions, enter plan mode first, present an implementation plan, and wait for
explicit 'proceed' or 'go ahead' before executing" so the PROACTIVELY behavior
and threshold (refer to the description text and the PROACTIVELY trigger
wording) are consistent.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 8138e091-881a-4d5d-8c74-1bdcf270d8ab
📒 Files selected for processing (18)
.claude-plugin/plugin.jsonREADME.mdagents/debugger.mdagents/orchestrator.mdcommands/develop.mdcommands/doctor.mddocs/context-loading.mddocs/cross-agent-workflows.mddocs/daily-habits.mddocs/new-features.mddocs/orchestration-patterns.mddocs/settings-guide.mdhooks/hooks.jsonreferences/claude-code-resources.mdsettings.example.jsonskills/context-optimizer/SKILL.mdskills/orchestrate/SKILL.mdskills/pro-workflow/SKILL.md
🚧 Files skipped from review as they are similar to previous changes (5)
- settings.example.json
- skills/context-optimizer/SKILL.md
- commands/doctor.md
- docs/cross-agent-workflows.md
- agents/debugger.md
Comment on lines
+2
to
+3
| description: Build a feature using Research > Plan > Implement phases | ||
| argument-hint: <feature description> |
There was a problem hiding this comment.
Frontmatter phase summary is outdated.
description lists only 3 phases, but the command defines 4 (Review & Commit is missing). Please align the summary string with the implemented workflow.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@commands/develop.md` around lines 2 - 3, Update the frontmatter 'description'
string to match the implemented workflow by listing all four phases (Research,
Plan, Implement, Review & Commit) instead of only three; edit the 'description'
value in the file so it explicitly includes "Research > Plan > Implement >
Review & Commit" (keep 'argument-hint' unchanged).
Comment on lines
+131
to
+132
| If you have many skills, they share a character budget for descriptions. Set `SLASH_COMMAND_TOOL_CHAR_BUDGET` env var to increase the limit if needed. | ||
|
|
There was a problem hiding this comment.
Clarify that character budget is an external runtime env var, not a pro-workflow config key.
This reads like a project setting; adding a short note (“set in your shell/Claude runtime env”) will prevent users from searching config.json for it.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/context-loading.md` around lines 131 - 132, The doc incorrectly implies
SLASH_COMMAND_TOOL_CHAR_BUDGET is a pro-workflow project config key; clarify
that SLASH_COMMAND_TOOL_CHAR_BUDGET is an external runtime environment variable
(set in your shell or the Claude/runtime environment), not a setting in
pro-workflow's config.json. Update the sentence around
SLASH_COMMAND_TOOL_CHAR_BUDGET to explicitly state it must be exported/defined
in your environment (e.g., shell or runtime env) and not placed in project
config files so readers won’t search for it in pro-workflow config.
| - `filesystem` - File ops | ||
| - `context7` — Live documentation lookup | ||
| - `playwright` — Browser automation (most token-efficient) | ||
| - `github` — PRs, issues, code search |
There was a problem hiding this comment.
Use official product casing: GitHub.
Replace `github` with `GitHub` for consistency with product naming.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@skills/pro-workflow/SKILL.md` at line 484, Update the product casing in the
markdown list item that currently contains the inline code token `github` to use
the official name `GitHub` (replace the token `` `github` `` with `` `GitHub`
``) so the entry reads "`GitHub` — PRs, issues, code search".
…itHub casing - Change "18 hook events" to "15 hook events" across README, plugin.json, marketplace.json, and SKILL.md to match hooks.json (3 events documented in new-features.md but not configured) - Add Alias column to Commands table documenting both namespaced (/pro-workflow:wrap-up) and short (/wrap-up) invocation forms - Add "Review & Commit" phase to /develop flow descriptions (4 phases, not 3) to match commands/develop.md - Fix **github** → **GitHub** product casing in MCP Config section
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
skills/pro-workflow/SKILL.md (1)
269-269:⚠️ Potential issue | 🟡 MinorCommand docs are internally inconsistent for
/learn.Line 269 tells users to run
/learn, but the command table omits it. Either add/learnto the table or update Line 269 to reference the current supported command path.Also applies to: 492-504
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@skills/pro-workflow/SKILL.md` at line 269, The README text "Run `/learn` for a topic-by-topic guide..." references a command that isn't listed in the command table; either add an entry for `/learn` to the command table (so it appears with its description and usage) or update that line to reference the actual supported command (e.g., replace `/learn` with the existing command name shown in the command table). Locate the exact string "Run `/learn` for a topic-by-topic guide..." and the command table block (the table around the commands previously noted at lines 492-504) and make the change so the doc and table are consistent.
♻️ Duplicate comments (2)
README.md (2)
183-183:⚠️ Potential issue | 🟡 Minor
/developphase text is still missing the final step.Line 183 truncates the flow at “Review” while other sections describe “Review & Commit.”
Suggested doc fix
-| `/pro-workflow:develop` | `/develop` | Multi-phase feature build (Research > Plan > Implement > Review) | +| `/pro-workflow:develop` | `/develop` | Multi-phase feature build (Research > Plan > Implement > Review & Commit) |🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` at line 183, Update the README table row that lists the `/pro-workflow:develop` and `/develop` entry so the phase description for `/develop` is corrected from ending with "Review" to "Review & Commit"; locate the table cell that currently reads "Multi-phase feature build (Research > Plan > Implement > Review)" and change it to "Multi-phase feature build (Research > Plan > Implement > Review & Commit)" to match other sections.
176-177:⚠️ Potential issue | 🟠 MajorQuick Start uses aliases as default despite alias prerequisites.
Lines 176-177 say short commands work only when aliases are defined, but Lines 386-389 instruct alias usage directly. This can break first-run onboarding.
Suggested doc fix
-# Then try it: -/develop add user authentication # Multi-phase feature build -/doctor # Check your setup -/wrap-up # End session properly +# Then try it (canonical plugin commands): +/pro-workflow:develop add user authentication # Multi-phase feature build +/pro-workflow:doctor # Check your setup +/pro-workflow:wrap-up # End session properly + +# Optional: short aliases (/develop, /doctor, /wrap-up) work if defined in ~/.claude/commands/Also applies to: 386-389
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@README.md` around lines 176 - 177, Update the Quick Start examples to avoid implying short aliases work by default: clarify that short commands (e.g., /wrap-up) only work if an alias is defined in ~/.claude/commands/, and change the examples at the lines referencing /pro-workflow:<name> and the instructions around the 386-389 alias usage to either show the full plugin prefix (/pro-workflow:<name>) as the default or explicitly instruct users to create the alias first; mention the exact symbols /pro-workflow:<name> and ~/.claude/commands/ so readers know which examples to follow or which file to edit.
🧹 Nitpick comments (1)
skills/pro-workflow/SKILL.md (1)
200-218: Add explicit "as of" date and direct link to official documentation for sustainability.Core claims are accurate per current Anthropic docs (1M context via
context-1m-2025-08-07beta header, Opus 4.6 and Sonnet 4.6 support, 200K default), but time-sensitive details merit an "as of [date]" note and a link to the Claude API Context Windows documentation to prevent future staleness as model capabilities evolve.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@skills/pro-workflow/SKILL.md` around lines 200 - 218, Add an "as of" date and a direct link to the official Claude Context Windows docs to the time-sensitive claims about Opus 4.6, Sonnet 4.6, and the 1M-token beta header (`context-1m-2025-08-07`) in SKILL.md and the suggested addition to CLAUDE.md: update the paragraph(s) mentioning "Opus 4.6", "Sonnet 4.6", and the 1M context beta to append "as of YYYY-MM-DD" and insert the URL https://docs.anthropic.com/en/docs/about-claude/models/overview (or a short linked reference) so readers can verify current capabilities; ensure the CLAUDE.md snippet under "Model Hints" also contains the same "as of" date and the documentation link.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@README.md`:
- Line 26: The table row showing "7 Reference Guides" doesn't match the six
listed items; update the README table row so count and items align by either
adding the missing guide name (e.g., "orchestration patterns") to the listed
items in the same <tr> that contains "7 Reference Guides" or change the count
label from "7 Reference Guides" to "6 Reference Guides"; locate the row
containing the exact text "7 Reference Guides" and adjust the count or append
the missing guide name to the second <td> accordingly.
In `@skills/pro-workflow/SKILL.md`:
- Line 3: Update the top-level description in SKILL.md where the description
line currently states "15 hooks" to reflect the new count "18 hooks" so the
metadata matches the PR's changes; specifically edit the description field text
(the line beginning with "description: Complete AI coding workflow system...")
to replace "15 hooks" with "18 hooks".
---
Outside diff comments:
In `@skills/pro-workflow/SKILL.md`:
- Line 269: The README text "Run `/learn` for a topic-by-topic guide..."
references a command that isn't listed in the command table; either add an entry
for `/learn` to the command table (so it appears with its description and usage)
or update that line to reference the actual supported command (e.g., replace
`/learn` with the existing command name shown in the command table). Locate the
exact string "Run `/learn` for a topic-by-topic guide..." and the command table
block (the table around the commands previously noted at lines 492-504) and make
the change so the doc and table are consistent.
---
Duplicate comments:
In `@README.md`:
- Line 183: Update the README table row that lists the `/pro-workflow:develop`
and `/develop` entry so the phase description for `/develop` is corrected from
ending with "Review" to "Review & Commit"; locate the table cell that currently
reads "Multi-phase feature build (Research > Plan > Implement > Review)" and
change it to "Multi-phase feature build (Research > Plan > Implement > Review &
Commit)" to match other sections.
- Around line 176-177: Update the Quick Start examples to avoid implying short
aliases work by default: clarify that short commands (e.g., /wrap-up) only work
if an alias is defined in ~/.claude/commands/, and change the examples at the
lines referencing /pro-workflow:<name> and the instructions around the 386-389
alias usage to either show the full plugin prefix (/pro-workflow:<name>) as the
default or explicitly instruct users to create the alias first; mention the
exact symbols /pro-workflow:<name> and ~/.claude/commands/ so readers know which
examples to follow or which file to edit.
---
Nitpick comments:
In `@skills/pro-workflow/SKILL.md`:
- Around line 200-218: Add an "as of" date and a direct link to the official
Claude Context Windows docs to the time-sensitive claims about Opus 4.6, Sonnet
4.6, and the 1M-token beta header (`context-1m-2025-08-07`) in SKILL.md and the
suggested addition to CLAUDE.md: update the paragraph(s) mentioning "Opus 4.6",
"Sonnet 4.6", and the 1M context beta to append "as of YYYY-MM-DD" and insert
the URL https://docs.anthropic.com/en/docs/about-claude/models/overview (or a
short linked reference) so readers can verify current capabilities; ensure the
CLAUDE.md snippet under "Model Hints" also contains the same "as of" date and
the documentation link.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: ca8dc3a0-e179-4a14-bf1c-381d33d8893f
📒 Files selected for processing (4)
.claude-plugin/marketplace.json.claude-plugin/plugin.jsonREADME.mdskills/pro-workflow/SKILL.md
🚧 Files skipped from review as they are similar to previous changes (1)
- .claude-plugin/marketplace.json
… dates - Add missing "orchestration patterns" to 7 Reference Guides table row - Revert hook count to 18 (covers all Claude Code events including 3 available but unconfigured: Setup, WorktreeCreate, WorktreeRemove) and add those 3 to the README hooks table with "(available, not configured)" notes - Fix /develop phase: "Review" → "Review & Commit" in Commands table - Clarify command invocation: plugin prefix /pro-workflow:<name> is default, short forms also work after install - Update Quick Start to use namespaced /pro-workflow:<name> commands - Add /learn command to SKILL.md command table with reference to commands/learn.md - Add "as of 2025-08" date and Anthropic docs link to model selection section and CLAUDE.md snippet for time-sensitive claims
This was referenced Mar 28, 2026
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
1 participant
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.
Summary
Major upgrade from v1.3 to v2.0. Adds comprehensive reference guides, orchestration patterns, cross-agent workflows, and new Claude Code features coverage. All original content informed by research across official docs, community workflows, and production use.
What's New
7 Reference Guides (all original):
docs/settings-guide.md— Complete settings hierarchy (5 levels), all permission modes, wildcard syntax, sandbox, env varsdocs/cli-cheatsheet.md— Every CLI flag, keyboard shortcut, 30+ slash commands, scripting patternsdocs/orchestration-patterns.md— Command > Agent > Skill architecture, 5 patterns, full frontmatter reference for commands/agents/skillsdocs/context-loading.md— How CLAUDE.md loads in monorepos (ancestor vs descendant), agent memory scopes, skills discovery, memory systems comparisondocs/cross-agent-workflows.md— Claude Code + Cursor + Codex + Gemini CLI config mapping, background agent patterns, MCP sharingdocs/new-features.md— Voice mode, agent teams, checkpointing/rewind, remote control, 10 new hook events, billing/usagedocs/daily-habits.md— Per-session/daily/weekly habits, debugging tips, terminal recommendations, anti-patterns2 New Agents:
orchestrator— Multi-phase development (Research > Plan > Implement) with confidence scoring and validation gatesdebugger— Systematic hypothesis-driven investigation with root cause analysis2 New Commands:
/develop— Multi-phase feature build with validation gates between phases/doctor— Health check for pro-workflow setup and Claude Code configuration2 New Skills:
context-optimizer— Token management, context budget planning, MCP auditorchestrate— Command > Agent > Skill wiring guide7 New Hook Events (18 total, up from 11):
Configuration:
settings.example.json— Production-ready template with permissions, spinner, output style, auto-compact at 80%mcp-config.example.json— Curated recommendations (context7, playwright, github) with scopes and approval settingsWhat's Updated
What's NOT in here
Test plan
/plugin install/developcommand with a multi-file feature/doctorhealth checkSummary by CodeRabbit
New Features
Documentation
Chores