Changes

• Send Accept: application/json, application/problem+json from the CodeAlive context-engine skill client so API ProblemDetails responses are negotiated consistently.
• Update chat continuation example to use the 24-character conversation ID returned by the API.
• Bump plugin version to 2.0.9.

Verification

• python3 -m pytest -q -> 18 passed.

What's Changed

codealive-context-engine skill

• SKILL.md: Tools table and "When to Use" section now clearly separate semantic search (default, by meaning) from grep search (exact text/regex). Tool reference descriptions updated.

• search.py: Docstring marks it as "the default discovery tool". Empty-result message guides agent to rephrase or try grep.py.

• grep.py: Docstring explains when to use exact text search. Empty-result message guides agent to check case or try search.py.

What's New

Windows / WSL Compatibility

• Add .gitattributes to force LF line endings for .sh files (prevents CRLF shebang breakage on WSL)
• Add WSL credential detection in check_auth.sh via cmdkey existence check
• Add WSL credential reading in api_client.py via powershell.exe P/Invoke (CredReadW)
• Add WSL credential read/store in setup.py with Windows-aware env var guidance

Skill Workflow Improvements

• Update skill guidance for search-first workflow
• Add canonical semantic and grep skill flows
• Add relationships drill-down and surface MCP triage hint
• Normalize self-hosted CodeAlive URLs in skills

Infrastructure

• Fix skills CI setup-python cache configuration
• Add automated runtime tests for CodeAlive skills
• Add Claude Desktop self-hosted best practices

Breaking Change

explore.py has been removed.

AI coding agents are better orchestrators than a hardcoded Python pipeline. Agents can:

• Inspect search results and apply freshness annotations
• Read() modified files between steps
• Decide whether to invoke chat.py based on search results
• Reason between each step of the workflow

Migration

Replace explore.py calls with the standard workflow:

1. search.py — find relevant code
2. Review results, Read() local files or fetch.py for external repos
3. Optionally chat.py for synthesized explanations

Other changes

• Updated SKILL.md, workflows.md, query-patterns.md
• Updated codealive-code-explorer agent rules
• Updated documentation site (skills.mdx)

Bug Fix

• Fix explore.py TypeError: All 5 exploration modes (understand, dependency, pattern, implement, debug) were passing an invalid include_content parameter to client.search(), causing a TypeError on every invocation. Removed the parameter from all calls.

What's New

• Search output improvements — snippet fallback shown when description is absent; misleading relevance score removed from output
• Real line numbers — fetch_artifacts now uses startLine from API response for accurate line numbers
• OpenClaw support — added to supported agents list and install paths table

Install / update

Plugin (Claude Code):

/plugin marketplace add CodeAlive-AI/codealive-skills
/plugin install codealive@codealive-marketplace

Skill (Claude Code, Cursor, Codex, OpenClaw, and 30+ others):

npx skills add CodeAlive-AI/codealive-skills@codealive-context-engine

What's New

• Fetch Artifacts tool — new fetch.py script to retrieve full source code content for search results from external repositories
• Description detail control — replaced --include-content with --description-detail short|full for finer control over search result verbosity
• Plugin manifest — added plugin.json for Anthropic plugin directory submission
• Plugin version bump — plugin.json version → 1.1.0
• Improved README — recommended plugin installation for Claude Code users

Bug Fixes

• Fixed duplicate skill registration when installed as both skill and plugin

Install / update

Plugin (Claude Code):

/plugin marketplace add CodeAlive-AI/codealive-skills
/plugin install codealive@codealive-marketplace

Skill (Claude Code, Cursor, Codex, etc.):

npx skills add CodeAlive-AI/codealive-skills@codealive-context-engine

CodeAlive Skills v1.0.0

Agent skills and Claude Code plugin for CodeAlive — semantic code search and AI-powered codebase Q&A across indexed repositories.

What's Included

Skill: codealive-context-engine

• Semantic search (search.py) — fast, low-cost code search across indexed repos
• Chat with codebase (chat.py) — synthesized AI answers about architecture and patterns
• Smart exploration (explore.py) — multi-step discovery workflows (understand, debug, implement, etc.)
• Data sources (datasources.py) — list indexed repositories and workspaces

Claude Code Plugin

• Authentication hook — warns about missing API key on session start
• codealive-code-explorer subagent — lightweight Haiku-based iterative code exploration that saves main conversation context

Authentication

• Interactive setup via python setup.py
• API key stored in OS credential store (macOS Keychain / Linux secret-tool / Windows Credential Manager)
• Clear error messages with setup instructions when key is missing

Installation

Universal (30+ agents):

npx skills add CodeAlive-AI/codealive-skills@codealive-context-engine

Claude Code plugin:

/plugin marketplace add CodeAlive-AI/codealive-skills
/plugin install codealive@codealive-marketplace

Setup:

python setup.py

Get your API key at app.codealive.ai/settings/api-keys.