All authors
sysevol-ai avatar

Claude Skills by sysevol-ai

github.com/sysevol-ai
14 skillsA× 140 installs1 views
Bm25 SearchA

<!-- SPDX-FileCopyrightText: 2025-2026 CodeNib Contributors

ai-agentsnode
0
86
Code To QueryA

<!-- SPDX-FileCopyrightText: 2025-2026 CodeNib Contributors SPDX-License-Identifier: Apache-2.0 --> Use to transform code snippets into search queries. Useful for finding similar code or understanding code context.

ai-agents
0
86
Codenib ContextA

Given the task/problem statement, it returns the relevant code in a single call: it searches for entry-point symbols (keyword + semantic) and **expands them along the call graph** (callers + callees), returning a compact, deduped, budget-capped set (name · file:line · kind · relation). No code bodies — `read` the few you want to confirm. This replaces the usual fan-out of repeated grep/search/read: one call gives you entry points *and* their structural neighbourhood, so you can answer or pinp...

ai-agents
0
86
Crossencoder RerankA

<!-- SPDX-FileCopyrightText: 2025-2026 CodeNib Contributors SPDX-License-Identifier: Apache-2.0 --> Use for fast, precise reranking after initial retrieval. Scores each (query, candidate) pair jointly using a cross-encoder model — more accurate than embedding similarity and faster than LLM listwise reranking. Best after bm25_search or embedding_search to refine the top results before generating an answer.

ai-agents
0
86
Embedding SearchA

<!-- SPDX-FileCopyrightText: 2025-2026 CodeNib Contributors

ai-agentsnode
0
86
Find CalleesA

"What does this function call?" — outgoing call-graph edges. Use to follow a symbol into its dependencies. Returns a compact list (name, file:line); call read for the bodies you care about.

ai-agentsgo
0
86
Find CallersA

"Who calls this function/class?" — incoming call-graph edges. Use for impact ("what breaks if I change X") and to walk a bug up into its caller. Returns a compact list (name, file:line); call read for the bodies you care about.

ai-agents
0
86
Hybrid SearchA

<!-- SPDX-FileCopyrightText: 2025-2026 CodeNib Contributors

ai-agentsnode
0
86
Llm RerankA

<!-- SPDX-FileCopyrightText: 2025-2026 CodeNib Contributors SPDX-License-Identifier: Apache-2.0 --> Use for high-quality reranking when precision matters most. Uses LLM to judge relevance. Slower but more accurate than embedding rerank. Best after initial retrieval to refine top results.

ai-agents
0
86
Lsp DefinitionA

LSP-compatible "go to definition" using the runtime's selected semantic provider. Use it when you have an exact source position from repository evidence and want the definition target without fanning out through raw grep. It deliberately uses the native LSP common denominator: a repository-relative file, 1-based line, and 0-based character. Use `lsp_route` for CodeNib-specific symbol-name navigation. Inputs are agent-facing: `line` is 1-based like `read` output. Results are compact node locat...

ai-agentsgonode
0
86
Lsp ReferencesA

LSP-compatible "find references" using the runtime's selected semantic provider. Use it when you need call/use sites for the symbol at an exact source position obtained from repository evidence. It deliberately uses the native LSP common denominator: a repository-relative file, 1-based line, and 0-based character. Use `lsp_route` for CodeNib-specific symbol-name navigation. Inputs are agent-facing: `line` is 1-based like `read` output. Results are compact definition/reference locations only; ...

ai-agents
0
86
Lsp RouteA

Static-index semantic route map backed by CodeNib's symbol graph, not a live LSP server. Use it early when a localization request names a symbol, error code, config field, API, or cross-file behavior and you need a compact map of likely endpoint, bridge/factory, provider/value, and type anchors. Call with the most specific symbol-like names you know: - `symbols=["NewResolver", "DefaultConfig"]` - `symbols=["F523"]` If no reliable symbol is known yet, call with `symbols=[]` plus the original r...

ai-agentsnodeapi
0
86
Repository SearchA

Search repository implementation with one focused query. The tool combines lexical and semantic rankings with reciprocal-rank fusion when both views are loaded, while reserving result slots for strong evidence unique to either branch; on a fast BM25-only index it uses lexical ranking without changing the interface. Every result includes a repository location and a bounded source excerpt focused on the query, so one large function cannot displace the rest of the evidence. Use this as the first...

ai-agentsdocumentation
0
86
TraceA

"How does X reach Y?" — the shortest call path between two symbols, following call edges grep can't (incl. indirect hops). Returns the ordered path (name, file:line per hop). Use to connect a symptom to its root cause across functions.

ai-agents
0
86