Install
$ agentstack add skill-tae2089-code-context-graph-ccg ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README — it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming — see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps — measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
ccg — Routing & Search
ccg is a Tree-sitter-based code graph tool. Complementary to Grep/Read, not a replacement. Choose based on task type.
Task Routing (most important)
| User intent | Tool | Why | | ------------------------------------------ | ----------------- | --------------------------------- | | "Where is X?" — simple location lookup | Grep + Read | Faster and cheaper than ccg | | "Find code related to X" — semantic search | ccg search | Annotation/keyword semantic match | | "What's affected if I change X?" | /ccg-analyze | Graph traversal | | "Understand structure/architecture" | /ccg-docs (RAG) | retrieve_docs first, then tree | | "Document intent/rules in code" | /ccg-annotate | AI annotation workflow | | "Manage multiple service codebases" | /ccg-namespace | MSA namespace isolation |
Don't use ccg when Grep is enough. ccg MCP context costs hundreds to thousands of tokens per task. For trivial tasks, it's pure overhead.
Core Commands
ccg build . # Build graph + search index (first time or after big changes)
ccg update . # Incremental — changed files only
ccg search "" # FTS search (includes annotations)
ccg status # Graph statistics
ccg docs --out docs # Generate docs + default RAG index
ccg serve # Start MCP server (stdio)
For remote or self-hosted MCP over Streamable HTTP, use ccg-server instead of ccg serve. Local ccg serve is stdio-only.
For detailed flags, use ccg --help or refer to MCP schema.
ccg search Patterns
Search by domain keywords (Korean works too). Annotation tags (@intent, @domainRule) are indexed alongside code.
ccg search "결제" # All payment-related functions (Korean)
ccg search "authentication" # Auth-related
ccg search --path internal/auth "login" # Path-scoped
Difference from Grep: Grep matches text, search matches semantic intent. Searching "결제" finds payment functions through their @intent 결제 처리 annotation.
Build Freshness
ccg build .fails with schema error → runccg migrate, retry- PostgreSQL or upgrading existing DB →
ccg migratefirst - Graph feels stale after code changes →
ccg update .
Core MCP Tools (commonly used)
| Tool | When | | ------------------ | -------------------------------------------- | | search | Semantic search | | query_graph | Structured queries (callers/callees/imports) | | get_node | Lookup by qualified name | | list_graph_stats | Graph size check | | get_minimal_context | Compact project snapshot and tool suggestions |
For other tools, see /ccg-analyze, /ccg-docs skills.
Agent Entry Pattern
For broad natural-language questions, start with generated docs/RAG before pulling exact graph edges:
ccg build .
ccg docs --out docs
Then use MCP retrieve_docs for bounded Markdown evidence, get_rag_tree to expand module context, and only then query_graph, get_node, or trace_flow for exact symbols and relationships.
Response Budget Rule
For LLM-agent use, prefer bounded graph queries. Start with limit=50 or limit=100 and follow has_more / next_offset rather than asking for a bulk result first.
Tools with explicit pagination:
| Tool | Parameters | | ---- | ---------- | | query_graph | limit, offset | | list_flows | limit, offset | | list_communities | limit, offset | | get_community | member_limit, member_offset when include_members=true | | get_architecture_overview | community_limit, community_offset, coupling_limit, coupling_offset |
High-volume analysis tools such as find_dead_code, find_suspect_fallback_edges, and broad architecture/onboarding prompts should be scoped by namespace, path, or a narrower first question before use.
Trade-offs (verified)
- Search tasks → ccg dominates (50–60% token reduction vs rg)
- Single location lookup → Grep+Read is cheaper (ccg is overhead)
- Miss-prevention matters (PR review, etc.) → ccg catches domain rules Grep misses
- Frequently changing code → factor in graph rebuild cost
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: tae2089
- Source: tae2089/code-context-graph
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.