skills.rest
standardbeagle
standardbeagleOFFICIAL
Explore all skills by this author →
WebsiteGitHub

lci-get-context

Maps lci's get_context and context manifest code paths, invariants, and competitor comparisons.

Updated Apr 27, 2026
Search Tags:#MCP#code search#code context#c++#call graph#get_context#context manifest
One-click install
npx skills add https://github.com/standardbeagle/lci-cpp --skill lci-get-context-standardbeagle
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: lci-get-context
Source: https://github.com/standardbeagle/lci-cpp/tree/main/.agents/skills/lci-get-context
Command: npx skills add https://github.com/standardbeagle/lci-cpp --skill lci-get-context-standardbeagle

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When working on lci's agent-facing context delivery (the MCP get_context tool and context manifest save/load), engineers need to locate the exact handlers, engine sections, and pinned Go-port behaviors without reading the whole C++ tree, and need grounded facts when comparing lci's context delivery against tools like Aider, Cursor, Serena, Cody, and Copilot. ## Core Features & Use Cases - Code map for get_context and context manifests: Entry points, ContextLookupEngine section fillers, base-63 object ID codec, manifest hydration with token budgets, and the LCF compact format, all with file:line references. - Port-trap and invariant catalog: Documents pinned stubs (fan_in, test_coverage, similar_objects), sorting/empty-list contracts, mode-vs-id-path behavior, and unwired propagator/annotator gaps so changes stay bug-for-bug compatible with the Go reference. - Probe recipes and product comparison: Ready-to-run JSON-RPC stdio probes, targeted gtest filters, and an honest axis-by-axis comparison of lci context delivery versus Aider, Cursor, Serena, Cody, and Copilot. - Use Case: You need to change how get_context renders source_excerpt without breaking golden tests; this Skill points you to attach_source_excerpt, the 12-line cap, the relevant fixtures, and the exact test filter to run. ## Quick Start Ask the AI to use the lci-get-context skill to explain how the get_context name path builds the seven-section CodeObjectContext and which pinned stubs must not change.

Frequently Asked Questions about lci-get-context

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find the get_context implementation in the lci codebase?▼

The MCP get_context handler lives in src/mcp/handlers_get_context.cpp (handle_get_context at :333), dispatched from handlers_core.cpp:322. The rich path constructs ContextLookupEngine from src/core/context_lookup.cpp, which fills seven sections in fixed order.

How does lci get_context compare to Aider, Serena, or Copilot context delivery?▼

lci delivers per-symbol packets with bounded 12-line excerpts and bare caller/callee names, while Aider uses a PageRanked repo-map, Serena uses LSP-backed find_symbol with bodies, and Cody/Copilot use embedding retrieval. The Skill provides an axis-by-axis table with verification pointers.

Why does get_context ignore the mode parameter when I pass an object id?▼

Mode is ignored on the id path by design: a request with id plus mode returns the compact envelope, verified by the GetContextModeWithIdFallsThroughToIdPath test. The rich context/metadata/performance envelope requires name plus mode or sections.

What are the pinned Go-port stubs in ContextLookupEngine?▼

Several fields are pinned for bug-for-bug Go parity: return_values and similar_objects are always empty, fan_in and call_frequency are always 0, test_coverage is false, requires_tests is true, and performance.component_breakdown is total divided by seven.

How do context manifests save and load token-budgeted code context?▼

The MCP context tool saves a 2-5 KB manifest of file/symbol refs with expansion directives, then load hydrates refs into source plus call-graph neighbors under context.max_tokens. Token accounting is bytes divided by 4, an estimate rather than a tokenizer count.

What are the limitations of lci get_context for citation-graded questions?▼

Callers, callees, and call_tree return bare names without locations, so citation-graded questions must use the callers tool instead. Additionally there are no embeddings, several sections are pinned stubs, and the semantic annotator is never wired into the production path.

Related Skills

Explore More Software Engineering Skills→
nixopusnixopus

cpp-deploy

Detect CMake or Meson configurations and build optimized C/C++ binaries with Ninja.

Official
Advanced
wandreandrade2018-makerwandreandrade2018-maker

V3 Performance Optimization

Benchmarks and validates claude-flow v3 performance targets including Flash Attention, HNSW search, and memory reduction.

Community
Advanced
wgordon17wgordon17

lsp-navigation

Navigate codebases with LSP for semantic symbol resolution and go-to-definition.

Community
Advanced
massgenmassgen
1.1k

serena

Locate symbol definitions and references across multi-file codebases using LSP-powered analysis.

Official
Advanced