weavatrix 1.17.4 → 1.17.6

package/README.md

@@ -12,11 +12,11 @@
 Weavatrix answers the questions agents otherwise invent: what breaks if this
 file changes, who consumes this n8n field, whether an MCP schema still
 accepts yesterday’s request, which viem call sites follow an ABI layout
-change, and whether `.weavatrix/architecture.json` still holds. **67
+change, and whether `.weavatrix/architecture.json` still holds. **64
 read-only operations**, one revision-bound graph, no LSP and no prompt pack.
 
 Codex, Claude Code, Cursor, and Grok get impact, architecture, APIs, Git
-history, duplicates, dead code, search, semantic links, temporal memory,
+history, duplicates, dead code, search, semantic links,
 exported n8n workflows (`n8n_*`), Dify YAML apps (`dify_*`), Agent
 Plugins/Skills/MCP catalogs (`agent_*`), Mermaid flowcharts (`diagram_*`),
 and Web3 ABI/consumer impact (`web3_*`). Those parsers are domains on
@@ -26,7 +26,9 @@ Weavatrix Core, not new MCP products. See
 
 It does not answer from a larger grep or an invented confidence score. Every
 bounded result can carry the repository revision, file, line, extractor,
-evidence kind, and confidence that produced it.
+evidence kind, and confidence that produced it. `search_code` keeps a
+resident content index for the process: the first search builds it, and
+later searches read it. A `glob` still walks files.
 
 This npm package is the convenient **prebuilt native tarball**. That is the
 only reason it exists on npm. It is not a TypeScript rewrite, not a Serena
@@ -58,7 +60,7 @@ weavatrix mcp .
 ```toml
 [mcp_servers.weavatrix]
 command = "npx"
-args = ["-y", "weavatrix@1.17.4", "mcp", "."]
+args = ["-y", "weavatrix@1.17.6", "mcp", "."]
 ```
 
 Pass an absolute repository path when the Codex process cwd is not the project
@@ -67,7 +69,7 @@ you intend to analyze.
 ### Claude Code
 
 ```sh
-claude mcp add weavatrix -- npx -y weavatrix@1.17.4 mcp .
+claude mcp add weavatrix -- npx -y weavatrix@1.17.6 mcp .
 ```
 
 ### Cursor
@@ -80,7 +82,7 @@ the plugin for the same server name:
   "mcpServers": {
     "weavatrix": {
       "command": "npx",
-      "args": ["-y", "weavatrix@1.17.4", "mcp", "${workspaceFolder}"]
+      "args": ["-y", "weavatrix@1.17.6", "mcp", "${workspaceFolder}", "--output-format=structured"]
     }
   }
 }
@@ -89,14 +91,14 @@ the plugin for the same server name:
 Profiles expose bounded views of the same engine:
 
 ```sh
-npx -y weavatrix@1.17.4 mcp . --profile=all
-npx -y weavatrix@1.17.4 mcp . --profile=code
-npx -y weavatrix@1.17.4 mcp . --profile=seo
-npx -y weavatrix@1.17.4 mcp . --profile=n8n
-npx -y weavatrix@1.17.4 mcp . --profile=dify
-npx -y weavatrix@1.17.4 mcp . --profile=agent
-npx -y weavatrix@1.17.4 mcp . --profile=diagram
-npx -y weavatrix@1.17.4 mcp . --profile=web3
+npx -y weavatrix@1.17.6 mcp . --profile=all
+npx -y weavatrix@1.17.6 mcp . --profile=code
+npx -y weavatrix@1.17.6 mcp . --profile=seo
+npx -y weavatrix@1.17.6 mcp . --profile=n8n
+npx -y weavatrix@1.17.6 mcp . --profile=dify
+npx -y weavatrix@1.17.6 mcp . --profile=agent
+npx -y weavatrix@1.17.6 mcp . --profile=diagram
+npx -y weavatrix@1.17.6 mcp . --profile=web3
 ```
 
 The package contains native binaries for Windows x64/arm64, macOS x64/arm64,
@@ -231,7 +233,7 @@ Three commits before `ec8bf30` this package was 1.9.2; the agent reads that
 follow-up to a diff without touching the worktree. Binary blobs fail closed
 instead of being decoded into garbage.
 
-## The 67 read-only operations
+## The 64 read-only operations
 
 | Workflow | Operations |
 | --- | --- |
@@ -243,7 +245,7 @@ instead of being decoded into garbage.
 | Architecture | `architecture_inventory`, `get_architecture_contract`, `verify_architecture`, `verify_capabilities`, `explain_architecture_violation`, `propose_architecture_exception` |
 | Local CI | `ci_restrictions`, `explain_restriction` |
 | Git and repositories | `git_history`, `git_read_blob`, `cross_repo_git`, `open_repo`, `list_known_repos`, `rebuild_graph` |
-| Native extensions | `vector_search`, `semantic_link`, `seo_link_suggestions`, `memory_context` |
+| Native extensions | `semantic_link` |
 | n8n workflows | `n8n_inventory`, `n8n_trace`, `n8n_context` |
 | Dify apps | `dify_inventory`, `dify_trace`, `dify_context` |
 | Agent packages | `agent_inventory`, `agent_trace`, `agent_context`, `agent_change_impact` |
@@ -260,6 +262,8 @@ organization, onion, ports-and-adapters, layered dependencies, and deployment
 uncertainty; the declared target style is never observation. The summary
 includes each hypothesis's status, supporting signals, contradictions, and
 unknowns; use `detail:"full"` for paged graph evidence. They
+recognize Rust `impl Trait for Type` as an explicit implementation signal,
+without mistaking inherent `impl Type` for one. They
 preserve nested Cargo, npm/TypeScript, Go, and Python components, many-to-many
 declared memberships, collision-free identities, configuration predicates,
 and immutable source evidence. Totals and SCC/cycle witnesses are computed
@@ -325,12 +329,12 @@ coding agent
     |
     | MCP over stdio
     v
-weavatrix 1.17.4
+weavatrix 1.17.6
     profile catalog · refresh · watcher · MCP framing
     |
     v
-weavatrix-rust 2.17.3
-    typed graph · analysis · 67 product operations including local CI, n8n, Dify, agent packages, Mermaid, and Web3
+weavatrix-rust 2.17.6
+    typed graph · analysis · 64 product operations including local CI, n8n, Dify, agent packages, Mermaid, and Web3
 ```
 
 This npm product owns MCP transport and native distribution. The

package/package.json

@@ -1,8 +1,8 @@
 {
   "name": "weavatrix",
-  "version": "1.17.4",
-  "weavatrixEngineVersion": "2.17.3",
-  "description": "Native repository intelligence for coding agents: 67 read-only MCP operations backed by typed evidence graphs, impact, architecture, local CI restrictions, APIs, Git, search, semantics, memory, n8n workflows, Dify apps, agent packages, Mermaid diagrams, and Web3 ABIs.",
+  "version": "1.17.6",
+  "weavatrixEngineVersion": "2.17.6",
+  "description": "Native repository intelligence for coding agents: 64 read-only MCP operations backed by typed evidence graphs, impact, architecture, local CI restrictions, APIs, Git, search, semantic links, n8n workflows, Dify apps, agent packages, Mermaid diagrams, and Web3 ABIs.",
   "author": "Sergii Ziborov <sergii.ziborov@gmail.com>",
   "license": "MIT",
   "homepage": "https://weavatrix.com",

package/server.json

@@ -2,19 +2,19 @@
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
   "name": "io.github.Weavatrix/weavatrix",
   "title": "Weavatrix",
-  "description": "Native Weavatrix MCP: 67 read-only tools for code, local CI, n8n, Dify, agents, Mermaid, Web3.",
+  "description": "Native Weavatrix MCP: 64 read-only tools for code, local CI, n8n, Dify, agents, Mermaid, Web3.",
   "repository": {
     "url": "https://github.com/Weavatrix/weavatrix",
     "source": "github"
   },
-  "version": "1.17.4",
+  "version": "1.17.6",
   "websiteUrl": "https://weavatrix.com",
   "packages": [
     {
       "registryType": "npm",
       "registryBaseUrl": "https://registry.npmjs.org",
       "identifier": "weavatrix",
-      "version": "1.17.4",
+      "version": "1.17.6",
       "runtimeHint": "npx",
       "transport": {
         "type": "stdio"

package/skill/SKILL.md

@@ -68,11 +68,43 @@ roots in their arguments. Multi-root retarget requires launching with
 `--allow-retarget`.
 
 `output_format` controls the MCP envelope, not the amount of detail: `text`
-still contains JSON. Choose each tool's summary/detail or result limits for
-conversational work. Use `token_budget` only as an additional ceiling; it can
-drop evidence arrays. Call `rebuild_graph`
-only when the repository changed before automatic refresh completed or when
-deliberately changing graph mode.
+is one JSON text block of the full payload, `json` mirrors that payload into
+text (about twice the bytes), and `structured` drops the mirror. Choose each
+tool's summary/detail or result limits for conversational work. Use
+`token_budget` only as an additional ceiling; it can drop evidence arrays.
+Call `rebuild_graph` only when the repository changed before automatic
+refresh completed or when deliberately changing graph mode.
+
+`search_code` reads a resident content index for this process. The first
+search builds it (`backend` is `liveindex`); later searches reuse it. A
+`glob` still walks the filesystem. It does not resolve symbols. Use
+`find_references`, `go_to_definition`, or `inspect_symbol` for a name.
+A column inside an identifier resolves; it does not have to be the first byte.
+
+`semantic_link` with a text `query` ranks symbols by the local deterministic
+token hash `weavatrix-token-hash-v1` (names, paths, and the declaration line).
+That is not a neural embedding, and Weavatrix does not create one. Pass
+`vectors` only when you already have embeddings. An empty `vectors` array
+without `query` is an error.
+
+`change_impact` with `files` or `target` reports the blast radius of those
+files even when the worktree matches HEAD. Those seeds use
+`change: "file_scope"` and `change_basis: "explicit_file"`. That is a
+hypothetical scope, not a measured diff. `status` stays COMPLETE when the
+tool ran. Read `evidence_completeness` and `stop_reason`. `MAX_NODES` means
+the combined cap omitted nodes.
+
+`git_history` without `path` is repository HEAD history. Pass `path` for the
+commits that touch one file or directory.
+
+`find_dead_code` treats every `main.go` and every Go `func main` as an entry.
+Receiver names and function-literal parameters are not symbols.
+
+`open_repo` is refused unless the process was started with `--allow-retarget`.
+A mismatched `expected_repository` means relaunch on that repository.
+
+`go.mod`, `Cargo.toml`, `package.json`, and `pyproject.toml` count as module
+markers. One marker leaves `modular_source` at `CANDIDATE`.
 
 Read [references/tool-routing.md](references/tool-routing.md) only when choosing
 among similar tools, tracing cross-repository transports, or interpreting

package/skill/references/tool-routing.md

@@ -20,7 +20,7 @@ cannot silently switch to another repository. Pass an explicit
 - [`list_communities`](tools/list_communities.md) — List deterministic coupling communities (containment and package edges excluded).
 - [`module_map`](tools/module_map.md) — Production folder map at a requested directory depth.
 - [`build_graph`](tools/build_graph.md) — Typed workspace, target, configuration, task, and runner topology from captured manifest evidence.
-- [`search_code`](tools/search_code.md) — Literal or Rust-regex repository search without ripgrep.
+- [`search_code`](tools/search_code.md) — Literal or Rust-regex search on the resident content index. A glob still walks files.
 - [`read_source`](tools/read_source.md) — Bounded source context by node or repository path.
 - [`inspect_symbol`](tools/inspect_symbol.md) — Definition, direct relationships and source evidence.
 - [`go_to_definition`](tools/go_to_definition.md) — Resolve the symbol at a source position to its definition without guessing by name.
@@ -69,12 +69,9 @@ cannot silently switch to another repository. Pass an explicit
 - [`list_known_repos`](tools/list_known_repos.md) — List repositories with an in-process graph.
 - [`rebuild_graph`](tools/rebuild_graph.md) — Rebuild the derived in-memory graph without source writes.
 
-## Optional supplied-data analysis
+## Semantic link
 
-- [`semantic_link`](tools/semantic_link.md) — Build inferred semantic graph evidence from supplied vectors.
-- [`vector_search`](tools/vector_search.md) — Exact or bounded approximate nearest-neighbor search.
-- [`seo_link_suggestions`](tools/seo_link_suggestions.md) — Directional SEO internal-link evidence from supplied page profiles.
-- [`memory_context`](tools/memory_context.md) — Compile bounded temporal memory context from supplied events.
+- [`semantic_link`](tools/semantic_link.md) — Rank a text query with the local token hash `weavatrix-token-hash-v1`, or link a non-empty caller-supplied vectors array. Nearest-neighbor search, SEO page links, and the event journal are library crates, not MCP tools.
 
 ## n8n workflows
 

package/skill/references/tools/agent_change_impact.md

@@ -18,7 +18,7 @@ Skill hits are declared `allowed-tools`, not proven calls.
 - `before` (string, required) — repository-relative previous catalog snapshot.
 - `after` (string, required) — repository-relative current catalog snapshot.
 - `max_results` (number, optional) — change page size, 1–500, default 100.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -33,5 +33,4 @@ Skill hits are declared `allowed-tools`, not proven calls.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/agent_context.md

@@ -13,7 +13,7 @@ Inspect one plugin, skill, or server after `agent_inventory` or `agent_trace`.
 - `label` (string, required) — exact node id or an unambiguous label.
 - `task` (string, optional) — what the caller intends to change or inspect.
 - `max_related` (number, optional) — relation page size, 1–200, default 24.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -28,5 +28,4 @@ Inspect one plugin, skill, or server after `agent_inventory` or `agent_trace`.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/agent_inventory.md

@@ -13,7 +13,7 @@ permissions.
 
 - `path` (string, optional) — repository-relative package file or path fragment.
 - `max_results` (number, optional) — inventory page size, 1–500, default 200.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -26,5 +26,4 @@ permissions.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/agent_trace.md

@@ -12,7 +12,7 @@ label can belong to two packages. Use an exact id if the label is ambiguous.
 
 - `label` (string, required) — exact node id or an unambiguous label.
 - `max_related` (number, optional) — relation page size, 1–200, default 48.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -26,5 +26,4 @@ label can belong to two packages. Use an exact id if the label is ambiguous.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/build_graph.md

@@ -13,7 +13,7 @@ with the analyzed snapshot rather than reread from a changing worktree.
 ## Inputs
 
 - `max_members` (integer, min 0).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -26,8 +26,7 @@ with the analyzed snapshot rather than reread from a changing worktree.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.
 
 `max_members` limits presentation only. Returned totals still describe the
 complete recovered model, and the result reports an incomplete page instead of

package/skill/references/tools/change_impact.md

@@ -4,7 +4,7 @@ Read-only Git change impact with graph evidence.
 
 ## When to use
 
-Map a working-tree diff, explicit files, or Git revisions onto affected graph nodes.
+Map a working-tree diff, explicit files, or Git revisions onto affected graph nodes. Naming `files` or `target` when that file matches HEAD still returns its symbols as `change: "file_scope"` and `change_basis: "explicit_file"`. That is a hypothetical blast radius, not a measured edit. A dirty file keeps the real symbol diff. `status` stays COMPLETE when the tool ran. Read `evidence_completeness` and `stop_reason`. `MAX_NODES` means the combined cap omitted nodes.
 
 Static impact is not proof that selected behavior executed.
 
@@ -16,9 +16,9 @@ Static impact is not proof that selected behavior executed.
 - `diff` (string).
 - `files` (array<string>).
 - `head_ref` (string).
-- `max_nodes` (integer, min 0).
+- `max_nodes` (integer, min 0, default 40). Combined cap on impacted nodes across every seed. Production callers are kept ahead of tests. A trimmed answer reports `MAX_NODES`.
 - `max_references` (integer, min 0).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 - `precision` ("graph").
 - `timeout_ms` (integer, min 0).
 
@@ -33,5 +33,4 @@ Static impact is not proof that selected behavior executed.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/context_bundle.md

@@ -15,7 +15,7 @@ Assemble a ranked, token-bounded workset around one task or target.
 - `max_references` (integer, min 0).
 - `max_related` (integer, min 0).
 - `max_source_files` (integer, min 0).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 - `precision` (string).
 - `timeout_ms` (integer, min 0).
 - `token_budget` (integer, min 1) — Approximate output ceiling in tokens (serialized bytes / 4); result arrays are trimmed from the tail to fit and the report states what was dropped.
@@ -32,5 +32,4 @@ Assemble a ranked, token-bounded workset around one task or target.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/cross_repo_git.md

@@ -15,7 +15,7 @@ Compare histories, shared commits, or diffs across explicitly supplied local rep
 - `head_ref` (string).
 - `left` (string).
 - `max_commits` (integer, min 0).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 - `revision` (string).
 - `right` (string).
 
@@ -31,5 +31,4 @@ Compare histories, shared commits, or diffs across explicitly supplied local rep
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/diagram_context.md

@@ -13,7 +13,7 @@ display name.
 - `label` (string, required) — graph id or unique native element label.
 - `task` (string, optional) — what the caller intends to change or inspect.
 - `max_related` (number, optional) — related row cap, 1–200, default 24.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -28,5 +28,4 @@ display name.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/diagram_inventory.md

@@ -12,7 +12,7 @@ Drawn arrows are declared architecture, not production calls.
 
 - `path` (string, optional) — repository-relative diagram file or path fragment.
 - `max_results` (number, optional) — inventory page size, 1–500, default 200.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -25,5 +25,4 @@ Drawn arrows are declared architecture, not production calls.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/diagram_trace.md

@@ -13,7 +13,7 @@ as a production `calls` edge.
 - `depth` (number, optional) — walk depth, 1–32, default 8.
 - `max_nodes` (number, optional) — page size, 1–500, default 100.
 - `cursor` (string, optional) — opaque page token from a previous page.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -27,5 +27,4 @@ as a production `calls` edge.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/dify_context.md

@@ -13,7 +13,7 @@ tasks. Secrets stay omitted.
 - `label` (string, required) — node, app, or variable id or graph label.
 - `task` (string, optional) — `explain` or `impact`. Impact adds consumer lists.
 - `max_related` (number, optional) — related-item page size.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -28,5 +28,4 @@ tasks. Secrets stay omitted.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/dify_inventory.md

@@ -11,7 +11,7 @@ not call a live Dify console API.
 
 - `path` (string, optional) — repository-relative YAML file or path fragment.
 - `max_results` (number, optional) — inventory page size, 1–500, default 200.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -24,5 +24,4 @@ not call a live Dify console API.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/dify_trace.md

@@ -14,7 +14,7 @@ or app. A DAG merge is a merge, not a cycle. Missing consumers stay missing.
 - `max_nodes` (number, optional) — page size, 1–500, default 100.
 - `cursor` (string, optional) — `v1:<offset>` or `v1:<offset>:<revision>`.
 - `direction` ("outgoing" | "incoming", optional) — walk direction.
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -28,5 +28,4 @@ or app. A DAG merge is a merge, not a cycle. Missing consumers stay missing.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/explain_architecture_violation.md

@@ -9,7 +9,7 @@ Retrieve bounded evidence for one active architecture-violation fingerprint.
 ## Inputs
 
 - `fingerprint` (string, required).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -23,5 +23,4 @@ Retrieve bounded evidence for one active architecture-violation fingerprint.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/find_dead_code.md

@@ -4,7 +4,7 @@ Conservative unreferenced-symbol review queue.
 
 ## When to use
 
-Build a conservative dead-code review queue with entry-point, test, config, dynamic, and external-use evidence.
+Build a conservative dead-code review queue. Every `main.go` and every Go `func main` is an entry, so `main` itself is reachable. Receiver names and function-literal parameters are not symbols. Exported symbols with no callers stay at confidence 50; private ones at 85. `REVIEW_ONLY` means the queue is a review list.
 
 Never delete from this result alone; confirm source and framework registration points.
 
@@ -14,7 +14,7 @@ Never delete from this result alone; confirm source and framework registration p
 - `include_tests` (boolean).
 - `kinds` (array<string>).
 - `min_confidence` (integer, min 0, max 100).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 - `path` (string).
 - `top_n` (integer, min 0).
 
@@ -29,5 +29,4 @@ Never delete from this result alone; confirm source and framework registration p
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/find_duplicates.md

@@ -16,7 +16,7 @@ Review bounded Type-1, Type-2, or Type-3 clone families before confirming candid
 - `min_similarity` (number, min 0, max 100) — 0..1 is a fraction; values above 1 through 100 are percentages.
 - `min_tokens` (integer, min 0).
 - `mode` ("strict" | "exact" | "renamed" | "near_miss").
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 - `top_n` (integer, min 0).
 
 ## Minimal call
@@ -30,5 +30,4 @@ Review bounded Type-1, Type-2, or Type-3 clone families before confirming candid
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.

package/skill/references/tools/find_references.md

@@ -16,7 +16,7 @@ List every recorded occurrence of one symbol after you have a label or a pinned
 - `scip_path` (string) — optional on-disk SCIP index; never spawned.
 - `max_results` (integer, min 1).
 - `cursor` (string).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 

package/skill/references/tools/get_architecture_contract.md

@@ -12,7 +12,7 @@ Read or preview the local target-architecture contract.
 - `baseline_mode` (string).
 - `candidate_contract` (object).
 - `confirm_token` (string).
-- `output_format` ("text" | "json" | "structured", default "json") — text returns the concise text block only; json returns structured output and mirrors it into text for clients that read only content; structured drops that mirror, which is the larger copy, and is safe only where the client reads structuredContent.
+- `output_format` ("text" | "json" | "structured", default "json") — text returns one JSON text block of the full payload, not a shorter summary; json returns structured output and mirrors that JSON into text, so the answer is about twice as large; structured drops the mirror and is the smaller payload when the client reads structuredContent.
 
 ## Minimal call
 
@@ -25,5 +25,4 @@ Read or preview the local target-architecture contract.
 }
 ```
 
-Use `output_format:"text"` for compact agent interaction. The live MCP
-`tools/list` schema remains authoritative for this installed version.
+Prefer `output_format:"structured"` when the client reads structuredContent. `text` is still the full JSON payload. The live MCP `tools/list` schema remains authoritative for this installed version.