Compare commits

..

13 Commits

Author SHA1 Message Date
Tobin South
95f822f892 Merge branch 'main' into add-plugin/pagerduty 2026-03-11 15:59:01 +00:00
Noah Zweben
bd041495bd update(plugin-json): point to the correct Semgrep plugin directory (#584)
The Semgrep plugin currently does not work correctly when used through
Claude because it is located within a subdirectory of the Semgrep
Marketplace repository. This issue was reported in:
https://github.com/anthropics/claude-plugins-official/issues/450

Previously, this could not be fixed due to a limitation in Claude Code's
handling of plugins located in subdirectories. Support for this was added
with the git-subdir feature, released in v2.1.69:
https://github.com/anthropics/claude-code/issues/30593

A fix for the Semgrep plugin was proposed once this version became the
latest release. Now that v2.1.69+ is available as latest, this PR
implements that fix.

https://claude.ai/code/cse_01RtW9KS12VZNFfWmWY6z9Pu

Co-authored-by: Claude <noreply@anthropic.com>
2026-03-10 15:59:33 -07:00
Daisy S. Hollman
00f13a5f46 Merge pull request #106 from obahareth/main
Add Ruby LSP plugin with inline lspServers configuration
2026-03-10 13:21:32 -07:00
Tobin South
7e94c732f6 Merge pull request #540 from anthropics/add-plugin/postman
add(plugin-json): postman
2026-03-10 17:12:50 +00:00
Tobin South
f2115dba9d Add pagerduty to marketplace 2026-03-09 10:44:19 +00:00
Tobin South
4fa27586e5 Add postman to marketplace 2026-03-06 10:48:54 -08:00
Noah Zweben
205b6e0b30 Update webhook closed PR message link to clau.de (#500)
Replace the Google Forms link with the new plugin directory
submission link (https://clau.de/plugin-directory-submission).

https://claude.ai/code/session_01NxRDJKDvFR2d4wC4ppDEDT

Co-authored-by: Claude <noreply@anthropic.com>
2026-03-03 17:08:52 -08:00
Sam Fishman
b7c995dd3c Merge pull request #496 from anthropics/sfishman/ralph-loop-tool-use-fix
fix(ralph-loop): stop hook fails on tool_use blocks; leaks across sessions
2026-03-03 16:35:13 -08:00
sfishman
028eccf544 address review: bound grep to tail -n 100; restore explicit error paths
- Split pipeline into two steps (extract lines, then parse) mirroring the
  original structure.
- set +e around the jq call so failures reach the $? check instead of
  aborting under set -e.
- The "no text content" branch remains removed (that was the original bug —
  all-tool-use turns now correctly yield empty text and the loop continues).
2026-03-04 00:25:42 +00:00
sfishman
8644df9ad5 fix(ralph-loop): isolate loop state to the session that started it
The state file lives at .claude/ralph-loop.local.md — project-scoped,
not session-scoped. The plugin's Stop hook fires in every Claude Code
session open in that project directory. So if session A starts a loop,
session B's Stop events also find the state file and block, feeding A's
prompt into B and consuming A's iteration budget.

This was masked by the transcript-parsing bug fixed in the previous
commit: that bug deleted the state file on the first Stop in any
session, so neither session looped. Fixing it exposed the leak.

Fix: setup writes CLAUDE_CODE_SESSION_ID into the frontmatter; the hook
compares against .session_id from its stdin JSON and exits silently on
mismatch. State files without session_id (written by old setup scripts)
fall through to preserve existing behavior.
2026-03-02 22:52:00 +00:00
sfishman
adfc379663 fix(ralph-loop): stop hook fails when last assistant block is tool_use
Claude Code writes each assistant content block (text/tool_use/thinking)
as its own JSONL line. The hook's `grep role:assistant | tail -1` would
grab whichever block happened to be last — often tool_use — then jq's
text filter returned empty string, triggering the 'no text content' path
which deletes the state file and exits without blocking.

Net effect: the loop silently never fires. In one observed session, 62%
of assistant lines were tool_use-only; the hook deleted state on the
very first Stop event every time.

Fix: slurp all assistant lines with jq -rs, flatten to text blocks only,
take the last. Empty result is now non-fatal — no text means no <promise>
tag, so the loop continues. Also absorbs jq parse errors (control chars
in text) via || fallback instead of aborting under set -e.
2026-03-02 20:39:50 +00:00
Kenshiro Nakagawa
55b58ec6e5 Merge pull request #457 from anthropics/kenshiro/export-plugins-20260224
chore(skill-creator): update to latest skill-creator
2026-02-24 20:28:20 -08:00
Omar Bahareth
80a2049c5d Add Ruby LSP plugin with inline lspServers configuration 2026-01-04 15:27:02 +03:00
5 changed files with 115 additions and 22 deletions

View File

@@ -251,6 +251,30 @@
}
}
},
{
"name": "ruby-lsp",
"description": "Ruby language server for code intelligence and analysis",
"version": "1.0.0",
"author": {
"name": "Anthropic",
"email": "support@anthropic.com"
},
"source": "./plugins/ruby-lsp",
"category": "development",
"strict": false,
"lspServers": {
"ruby-lsp": {
"command": "ruby-lsp",
"extensionToLanguage": {
".rb": "ruby",
".rake": "ruby",
".gemspec": "ruby",
".ru": "ruby",
".erb": "erb"
}
}
}
},
{
"name": "agent-sdk-dev",
"description": "Development kit for working with the Claude Agent SDK",
@@ -680,10 +704,33 @@
"description": "Semgrep catches security vulnerabilities in real-time and guides Claude to write secure code from the start.",
"category": "security",
"source": {
"source": "url",
"url": "https://github.com/semgrep/mcp-marketplace.git"
"source": "git-subdir",
"url": "https://github.com/semgrep/mcp-marketplace.git",
"path": "plugin"
},
"homepage": "https://github.com/semgrep/mcp-marketplace.git"
},
{
"name": "pagerduty",
"description": "Enhance code quality and security through PagerDuty risk scoring and incident correlation. Score pre-commit diffs against historical incident data and surface deployment risk before you ship.",
"category": "monitoring",
"source": {
"source": "url",
"url": "https://github.com/PagerDuty/claude-code-plugins.git",
"sha": "b16c23e0d790deceaa7a6182616d0e36673f2eae"
},
"homepage": "https://github.com/PagerDuty/claude-code-plugins"
},
{
"name": "postman",
"description": "Full API lifecycle management for Claude Code. Sync collections, generate client code, discover APIs, run tests, create mocks, publish docs, and audit security. Powered by the Postman MCP Server.",
"category": "development",
"source": {
"source": "url",
"url": "https://github.com/Postman-Devrel/postman-claude-code-plugin.git",
"sha": "0714280351c1a137e79aad465a66730511ffbd57"
},
"homepage": "https://learning.postman.com/docs/developer/postman-mcp-server/"
}
]
}

View File

@@ -36,7 +36,7 @@ jobs:
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: context.payload.pull_request.number,
body: `Thanks for your interest! This repo only accepts contributions from Anthropic team members. If you'd like to submit a plugin to the marketplace, please submit your plugin [here](https://docs.google.com/forms/d/e/1FAIpQLSdeFthxvjOXUjxg1i3KrOOkEPDJtn71XC-KjmQlxNP63xYydg/viewform).`
body: `Thanks for your interest! This repo only accepts contributions from Anthropic team members. If you'd like to submit a plugin to the marketplace, please submit your plugin [here](https://clau.de/plugin-directory-submission).`
});
await github.rest.pulls.update({

View File

@@ -24,6 +24,16 @@ MAX_ITERATIONS=$(echo "$FRONTMATTER" | grep '^max_iterations:' | sed 's/max_iter
# Extract completion_promise and strip surrounding quotes if present
COMPLETION_PROMISE=$(echo "$FRONTMATTER" | grep '^completion_promise:' | sed 's/completion_promise: *//' | sed 's/^"\(.*\)"$/\1/')
# Session isolation: the state file is project-scoped, but the Stop hook
# fires in every Claude Code session in that project. If another session
# started the loop, this session must not block (or touch the state file).
# Legacy state files without session_id fall through (preserves old behavior).
STATE_SESSION=$(echo "$FRONTMATTER" | grep '^session_id:' | sed 's/session_id: *//' || true)
HOOK_SESSION=$(echo "$HOOK_INPUT" | jq -r '.session_id // ""')
if [[ -n "$STATE_SESSION" ]] && [[ "$STATE_SESSION" != "$HOOK_SESSION" ]]; then
exit 0
fi
# Validate numeric fields before arithmetic operations
if [[ ! "$ITERATION" =~ ^[0-9]+$ ]]; then
echo "⚠️ Ralph loop: State file corrupted" >&2
@@ -77,35 +87,39 @@ if ! grep -q '"role":"assistant"' "$TRANSCRIPT_PATH"; then
exit 0
fi
# Extract last assistant message with explicit error handling
LAST_LINE=$(grep '"role":"assistant"' "$TRANSCRIPT_PATH" | tail -1)
if [[ -z "$LAST_LINE" ]]; then
echo "⚠️ Ralph loop: Failed to extract last assistant message" >&2
# Extract the most recent assistant text block.
#
# Claude Code writes each content block (text/tool_use/thinking) as its own
# JSONL line, all with role=assistant. So slurp the last N assistant lines,
# flatten to text blocks only, and take the last one.
#
# Capped at the last 100 assistant lines to keep jq's slurp input bounded
# for long-running sessions.
LAST_LINES=$(grep '"role":"assistant"' "$TRANSCRIPT_PATH" | tail -n 100)
if [[ -z "$LAST_LINES" ]]; then
echo "⚠️ Ralph loop: Failed to extract assistant messages" >&2
echo " Ralph loop is stopping." >&2
rm "$RALPH_STATE_FILE"
exit 0
fi
# Parse JSON with error handling
LAST_OUTPUT=$(echo "$LAST_LINE" | jq -r '
.message.content |
map(select(.type == "text")) |
map(.text) |
join("\n")
# Parse the recent lines and pull out the final text block.
# `last // ""` yields empty string when no text blocks exist (e.g. a turn
# that is all tool calls). That's fine: empty text means no <promise> tag,
# so the loop simply continues.
# (Briefly disable errexit so a jq failure can be caught by the $? check.)
set +e
LAST_OUTPUT=$(echo "$LAST_LINES" | jq -rs '
map(.message.content[]? | select(.type == "text") | .text) | last // ""
' 2>&1)
JQ_EXIT=$?
set -e
# Check if jq succeeded
if [[ $? -ne 0 ]]; then
if [[ $JQ_EXIT -ne 0 ]]; then
echo "⚠️ Ralph loop: Failed to parse assistant message JSON" >&2
echo " Error: $LAST_OUTPUT" >&2
echo " This may indicate a transcript format issue" >&2
echo " Ralph loop is stopping." >&2
rm "$RALPH_STATE_FILE"
exit 0
fi
if [[ -z "$LAST_OUTPUT" ]]; then
echo "⚠️ Ralph loop: Assistant message contained no text content" >&2
echo " This may indicate a transcript format issue." >&2
echo " Ralph loop is stopping." >&2
rm "$RALPH_STATE_FILE"
exit 0

View File

@@ -141,6 +141,7 @@ cat > .claude/ralph-loop.local.md <<EOF
---
active: true
iteration: 1
session_id: ${CLAUDE_CODE_SESSION_ID:-}
max_iterations: $MAX_ITERATIONS
completion_promise: $COMPLETION_PROMISE_YAML
started_at: "$(date -u +%Y-%m-%dT%H:%M:%SZ)"

View File

@@ -0,0 +1,31 @@
# ruby-lsp
Ruby language server for Claude Code, providing code intelligence and analysis.
## Supported Extensions
`.rb`, `.rake`, `.gemspec`, `.ru`, `.erb`
## Installation
### Via gem (recommended)
```bash
gem install ruby-lsp
```
### Via Bundler
Add to your Gemfile:
```ruby
gem 'ruby-lsp', group: :development
```
Then run:
```bash
bundle install
```
## Requirements
- Ruby 3.0 or later
## More Information
- [Ruby LSP Website](https://shopify.github.io/ruby-lsp/)
- [GitHub Repository](https://github.com/Shopify/ruby-lsp)