token-goat 2.9.6 → 2.9.9

package/README.md

@@ -37,7 +37,7 @@ Restart your AI sessions. Run `token-goat stats` a couple of minutes after your
 
 > **Built and continually improved, free, by one person. If it saves you tokens, drop a ⭐️ at the top of this page. One click. Makes my day. Also, if you'd like anything added, [drop me a line](mailto:token-goat@dfkhelper.com).**
 
-[Install](docs/install.md) · [CLI](docs/cli.md) · [What gets installed?](docs/install.md#what-gets-installed) · [Stats](#stats-display) · [Security & uninstall](docs/security.md)
+[Install](docs/install.md) · [CLI](docs/cli.md) · [Architecture Q&A](docs/architecture-qa.html) · [What gets installed?](docs/install.md#what-gets-installed) · [Stats](#stats-display) · [Security & uninstall](docs/security.md)
 
 ---
 
@@ -129,7 +129,7 @@ The fastest way to reduce AI token costs is fixing these five, not writing short
 | `claude-in-chrome`'s `computer`/`browser_batch` return a raw, full-resolution base64 screenshot in-band, with no destination-file option to redirect through image-shrink | Inline screenshot blocks are shrunk via the same image-shrink pipeline in place, and a repeated `Tab Context:` listing (appended to nearly every call, often unchanged) collapses to a placeholder once seen unchanged this session |
 | Agent tool spawns a subagent with no orientation and no reuse hints | A `PreToolUse` handler appends a compact briefing pack to the prompt: a one-line project-map summary, 2-3 recent cached-output IDs, and a surgical-read reminder (~300 tokens) |
 | Subagent's own final report runs long and gets discarded once the parent moves on | Agent tool results ≥8000 characters get a recall pointer appended (`token-goat recall`); the original report always reaches the parent untouched |
-| Agent tool spawns with no `subagent_type` (or an explicit `general-purpose`), so the lane starts by paying for every tool and MCP schema on the machine | Once per session, a `PostToolUse` advisory names up to three `tools:`-restricted agent definitions found in `~/.claude/agents` and suggests passing one as `subagent_type` on a future spawn that fits. It fires only when at least one restricted definition exists, states outright that the observed spawn already ran and nothing was saved, and records a zero-credit stat |
+| Agent tool spawns with no `subagent_type` (or an explicit `general-purpose`), so the lane starts by paying for every tool and MCP schema on the machine | Once per session, a `PostToolUse` advisory names up to three `tools:`-restricted agent definitions found in `~/.claude/agents` or the current project's `.claude/agents` and suggests passing one as `subagent_type` on a future spawn that fits. It fires only when at least one restricted definition exists, states outright that the observed spawn already ran and nothing was saved, and records a zero-credit stat |
 | Large MCP tool result (≥2 KB) is a homogeneous array of objects, e.g. a list/search result | Deterministic structural compression: table-ified into one header row + tab-delimited rows, with columns constant across every row hoisted into a single `constant:` line instead of repeated per row; only applied when it saves ≥15%. Full original always recoverable via `token-goat bash-output <id>` (labeled `[token-goat: compressed, full via mcp-output <id>]`). Disable with `TOKEN_GOAT_MCP_COMPRESS=0` |
 | Large MCP tool result (≥2 KB) doesn't table-ify (an object-rooted config dump, a single-resource response, a nested settings tree) but most of its bytes are `null`/`""`/`[]`/`{}` | Deterministic empty-value pruning: recursively drops those four empty shapes to a fixed point (a container left empty by its own dropped children is dropped too) and renders the remainder as compact JSON with a trailing `dropped N empty value(s) (null, "", [], {})` summary line; `0` and `false` are never dropped; only applied when it saves ≥15%. Same full-recovery-by-id guarantee and `TOKEN_GOAT_MCP_COMPRESS=0` opt-out as the table pass |
 | GitHub MCP tool result (`list_pull_requests`, `list_issues`, `search_code`, `get_file_contents`, `pull_request_read`, …) carries dozens of boilerplate fields per object | GitHub compression pack strips `_links`, `node_id`, `gravatar_id`, `site_admin`, and every `*_url` field (`avatar_url`, `html_url`, `events_url`, `gists_url`, `followers_url`, …) except `download_url`/`git_url`/`clone_url`/`ssh_url`, before handing the shrunk JSON to the same table-ifying pass — same `TOKEN_GOAT_MCP_COMPRESS=0` opt-out and full-recovery-by-id guarantee |
@@ -169,6 +169,9 @@ The fastest way to reduce AI token costs is fixing these five, not writing short
 | Fetched web content lands raw in model context | Always wrapped in an untrusted-content fence; scanned for attack patterns, and a matched pattern name is written to the log and into the fence's notice |
 | Chatty log repeats the same error or event thousands of times | `token-goat logfold` collapses consecutive duplicates to `[Nx]` counts; same event logged with different timestamps or request IDs folds correctly — ~90–95% smaller on repetitive logs |
 | Reading poetry.lock or package-lock.json to find a pinned version | `token-goat lockdeps` returns a name/version table of direct dependencies; optional packages and transitive entries excluded |
+| Large SVG / diagram file (≥8 KB) read in full | Coordinate path flooding blocked; extracted layer/group IDs and title shown; redirects to `token-goat xml-outline` and `xml-query` |
+| Broad recursive Glob sweep (`*`, `**/*`) on root directory | Pre-Glob hook warns against tree-dumping and points at `token-goat map --compact` for fast, lightweight structure inspection |
+| Compound test/build pipeline (`npm run build && npm run typecheck && npm test`) | Post-Bash hook routes chained build/test/lint commands to `generic-ci` compression, dropping verbose passing steps and compiler noise |
 
 On a per-token API plan, 100K wasted tokens per session runs about $0.30. Five sessions a week is ~$450/year. AI coding cost reduction at that scale comes from fixing the waste, not from using the product less. Token-goat is free. And on subscription plans, it can result in limits feeling 10x higher.
 
@@ -320,7 +323,7 @@ corpus of captured shell output through the real compression path and reports ho
 result got, alongside a count of the lines each case is not allowed to lose. Deleting output makes
 the first number look better, so the second one is what keeps the first honest.
 
-Every command, with its flags, its output, and the comparison workflows that chain them: **[CLI reference](docs/cli.md)**.
+Every command, with its flags, its output, and the comparison workflows that chain them: **[CLI reference](docs/cli.md)**. For architectural analysis, empirical evaluations, and technical answers to common engineering questions: **[Architecture & Technical Q&A](docs/architecture-qa.html)**.
 
 ## MCP server
 
@@ -422,7 +425,7 @@ never drift on where `mcp.json` lives or what key name it looks for.
 
 **What the index actually holds, in plain terms.** The point of a surgical read is returning a function body without the file around it, which means the database stores those bodies. `symbols.body` holds the source text of every indexed symbol, `symbols.docstring` its doc comment, `refs.context` the line around each reference, and `chunks.text` the passages that semantic search embeds. There is also a full-text index over the bodies and docstrings. So the database is not a list of names and line numbers: it is a substantial copy of your source, sitting in a plain unencrypted SQLite file outside the repository.
 
-The file-by-file table for each harness, and the path that file sits at: **[What gets installed](docs/install.md#what-gets-installed)**.
+The file-by-file table for each harness, and the path that file sits at: **[What gets installed](docs/install.md#what-gets-installed)** (see also **[Permissions & auto-approval](docs/install.md#command-auto-approval-and-permissions)**).
 
 ## Zero maintenance
 

package/SECURITY.md

@@ -101,7 +101,7 @@ For a scanner that ingests a bill of materials rather than a lockfile, `npm run
 
 ## Verifying what you installed
 
-Every published version is built and pushed by one workflow, [`.github/workflows/publish.yml`](.github/workflows/publish.yml), which runs only when a GitHub release is published (or manually, and then only from `main`). It publishes with npm provenance, so npm holds a signed attestation tying the tarball to the commit and workflow run that produced it. Nothing is ever published from a laptop.
+Every published version is built and pushed by one workflow, [`.github/workflows/publish.yml`](.github/workflows/publish.yml), which runs only when a GitHub release is published (or manually, and then only from `main`). It publishes with npm provenance, so npm holds a signed attestation tying the tarball to the commit and workflow run that produced it. Nothing is ever published from a developer workstation.
 
 To check a copy you already have:
 

package/dist/{token-goat-chunk-3HJQR4OO.mjs → token-goat-chunk-BYDDWQ2T.mjs}

@@ -47,6 +47,7 @@ import {
   getFilePath,
   getHarnessName,
   globalDbPath,
+  hasUnquotedOperator,
   isAutoTriggerMultiplierExplicit,
   isCaseInsensitiveFs,
   isCodeFenceDelimiter,
@@ -100,7 +101,7 @@ import {
   toDisplayPath,
   toKB,
   withFileLock
-} from "./token-goat-chunk-T2OE7MYM.mjs";
+} from "./token-goat-chunk-KIEOFWLL.mjs";
 import {
   registerReset
 } from "./token-goat-chunk-EEIDFMEM.mjs";
@@ -6688,11 +6689,15 @@ function getFileLineRanges(filePath) {
   return _fileLineRanges.get(foldPath(normalizePath(filePath))) ?? [];
 }
 var MAX_SERVED_OUTPUTS_PER_FILE = 8;
+var GENERIC_SERVED_OUTPUT_KEY = "\0<bash-generic-served>";
+var GENERIC_SERVED_OUTPUT_FOLDED_KEY = foldPath(normalizePath(GENERIC_SERVED_OUTPUT_KEY));
+var MAX_GENERIC_SERVED_OUTPUTS = 32;
 function recordFileServedOutput(filePath, outputId) {
   const key = foldPath(normalizePath(filePath));
+  const cap = key === GENERIC_SERVED_OUTPUT_FOLDED_KEY ? MAX_GENERIC_SERVED_OUTPUTS : MAX_SERVED_OUTPUTS_PER_FILE;
   const ids = (_fileServedOutputs.get(key) ?? []).filter((id) => id !== outputId);
   ids.push(outputId);
-  if (ids.length > MAX_SERVED_OUTPUTS_PER_FILE) ids.splice(0, ids.length - MAX_SERVED_OUTPUTS_PER_FILE);
+  if (ids.length > cap) ids.splice(0, ids.length - cap);
   _fileServedOutputs.set(key, ids);
 }
 function getFileServedOutputs(filePath) {
@@ -7624,6 +7629,7 @@ function isPsMultilineSystemQuery(cmd) {
 }
 function getMonitoringRecallHint(cmd) {
   const trimmed = cmd.trim();
+  if (hasUnquotedOperator(trimmed, ["&&", "||", ";"])) return null;
   for (const { pattern, recallHint } of MONITORING_COMMAND_PATTERNS) {
     if (pattern.test(trimmed)) return recallHint;
   }
@@ -8191,13 +8197,15 @@ function mergeLineRanges(disk, mem) {
   }
   return Array.from(byPath.entries());
 }
+var GENERIC_SERVED_OUTPUT_FOLDED_KEY2 = foldPath(normalizePath(GENERIC_SERVED_OUTPUT_KEY));
 function mergeServedOutputs(disk, mem) {
   const byPath = /* @__PURE__ */ new Map();
   for (const [filePath, ids] of disk) byPath.set(filePath, [...ids]);
   for (const [filePath, ids] of mem) {
     const prev = byPath.get(filePath) ?? [];
     const merged = [...prev.filter((id) => !ids.includes(id)), ...ids];
-    if (merged.length > MAX_SERVED_OUTPUTS_PER_FILE) merged.splice(0, merged.length - MAX_SERVED_OUTPUTS_PER_FILE);
+    const cap = filePath === GENERIC_SERVED_OUTPUT_FOLDED_KEY2 ? MAX_GENERIC_SERVED_OUTPUTS : MAX_SERVED_OUTPUTS_PER_FILE;
+    if (merged.length > cap) merged.splice(0, merged.length - cap);
     byPath.set(filePath, merged);
   }
   return Array.from(byPath.entries());
@@ -9939,7 +9947,7 @@ function pruneShrinkCache() {
   }
 }
 async function finalizeShrinkResult(result, filePath) {
-  const basename12 = path6.basename(filePath);
+  const basename12 = displaySafePath(path6.basename(filePath));
   const shrinkSaved = result.originalBytes - result.shrunkBytes;
   const tier = loadConfig().image_shrink.vision_tier;
   recordStat("image_shrink", shrinkSaved, visionTokensSaved(result.originalWidth, result.originalHeight, result.width, result.height, tier), void 0, basename12);
@@ -10260,6 +10268,47 @@ function modelDir() {
 function downloadUrl(file) {
   return `https://huggingface.co/${DEFAULT_MODEL}/resolve/${PINNED_MODEL_REVISION}/${file.name}`;
 }
+function sharedModelCacheDir() {
+  const raw = process.env["TOKEN_GOAT_MODEL_CACHE_DIR"]?.trim();
+  if (!raw) return null;
+  return path7.join(raw, ...DEFAULT_MODEL.split("/"), PINNED_MODEL_REVISION);
+}
+async function copyFromSharedCache(shared, file, target) {
+  const source = path7.join(shared, file.name);
+  const temp = `${target}.${process.pid}.shared`;
+  try {
+    if (!fs11.existsSync(source)) return false;
+    fs11.copyFileSync(source, temp);
+    if (await sha256Of(temp) !== file.sha256) {
+      fs11.rmSync(temp, { force: true, maxRetries: 20, retryDelay: 25 });
+      fs11.rmSync(source, { force: true, maxRetries: 20, retryDelay: 25 });
+      return false;
+    }
+    fs11.renameSync(temp, target);
+    return true;
+  } catch {
+    try {
+      fs11.rmSync(temp, { force: true, maxRetries: 20, retryDelay: 25 });
+    } catch {
+    }
+    return false;
+  }
+}
+function publishToSharedCache(shared, file, target) {
+  const destination = path7.join(shared, file.name);
+  const temp = `${destination}.${process.pid}.partial`;
+  try {
+    if (fs11.existsSync(destination)) return;
+    fs11.mkdirSync(path7.dirname(destination), { recursive: true });
+    fs11.copyFileSync(target, temp);
+    fs11.renameSync(temp, destination);
+  } catch {
+    try {
+      fs11.rmSync(temp, { force: true, maxRetries: 20, retryDelay: 25 });
+    } catch {
+    }
+  }
+}
 var _ort = null;
 var _ortError = null;
 var _ortLoadAttempted = false;
@@ -10355,6 +10404,7 @@ async function ensureModelFiles(modelName = DEFAULT_MODEL) {
   ensureDataDirPrivate();
   const dir = modelDir();
   const offline = loadConfig().network.offline;
+  const shared = sharedModelCacheDir();
   for (const file of MODEL_FILES) {
     const target = path7.join(dir, file.name);
     if (fs11.existsSync(target)) {
@@ -10362,6 +10412,10 @@ async function ensureModelFiles(modelName = DEFAULT_MODEL) {
       if (digest === file.sha256) continue;
       fs11.rmSync(target, { force: true });
     }
+    if (shared) {
+      fs11.mkdirSync(path7.dirname(target), { recursive: true });
+      if (await copyFromSharedCache(shared, file, target)) continue;
+    }
     if (offline) {
       throw new Error(
         `Offline mode is on (network.offline): refusing to download ${file.name} for the embedding model. Copy the pinned files into ${dir} on a connected machine to use semantic search here.`
@@ -10372,6 +10426,7 @@ async function ensureModelFiles(modelName = DEFAULT_MODEL) {
       `Downloading the embedding model, once (${file.name}, ${Math.round(file.bytes / 1024 / 1024)} MB) into ${dir}`
     );
     await download(file, target);
+    if (shared) publishToSharedCache(shared, file, target);
   }
   return dir;
 }
@@ -11061,7 +11116,7 @@ function _pathPriorityPenalty(filePath) {
 
 // src/parser_fingerprint.ts
 init_define_import_meta_env();
-var PARSER_FINGERPRINT = "567346f98dadd8cb";
+var PARSER_FINGERPRINT = "d96c01d5957651c0";
 
 // src/parser.ts
 init_define_import_meta_env();
@@ -14798,16 +14853,26 @@ ${commented}`);
 // src/parser.ts
 var _require3 = createRequire3(import.meta.url);
 var _parserCtor;
+var _parserCtorError = null;
+var _parserCtorOverride = void 0;
 var _grammarCache = /* @__PURE__ */ new Map();
 function loadParserCtor() {
+  if (_parserCtorOverride !== void 0) return _parserCtorOverride;
   if (_parserCtor !== void 0) return _parserCtor;
   try {
     _parserCtor = _require3("tree-sitter");
-  } catch {
+  } catch (e) {
     _parserCtor = null;
+    _parserCtorError = e instanceof Error ? e : new Error(String(e));
   }
   return _parserCtor;
 }
+function treeSitterCoreAvailable() {
+  return loadParserCtor() !== null;
+}
+function treeSitterCoreLoadError() {
+  return _parserCtorError;
+}
 var CPP_HEADER_SNIFF_RE = /\bclass\s+\w|\bnamespace\s+\w|\btemplate\s*<|::\s*\w|\b(?:public|private|protected)\s*:/;
 var MAX_SYMBOL_BODY_CHARS = SYMBOL_BODY_CHAR_CAP;
 function boundSymbolBody(body) {
@@ -18768,6 +18833,9 @@ var FILE_TYPE_THRESHOLDS = {
   // always intercept (any size)
   html: 5e4,
   txt: 2e4,
+  log: 1e4,
+  svg: 8e3,
+  xml: 2e4,
   csv: 1e4,
   tsv: 1e4,
   transcript: 1e4,
@@ -18835,8 +18903,9 @@ ${headings.join("\n")}` : ""].filter(Boolean).join("\n")
 }
 function handleTxt(filePath, content, contentLengthHint) {
   const length = contentLengthHint ?? content.length;
-  if (length < FILE_TYPE_THRESHOLDS.txt) return { shouldBlock: false, message: "" };
   const isLog = /\.(log|out|err|trace)$/i.test(filePath) || /[\\/]logs[\\/]/.test(filePath);
+  const threshold = isLog ? FILE_TYPE_THRESHOLDS.log : FILE_TYPE_THRESHOLDS.txt;
+  if (length < threshold) return { shouldBlock: false, message: "" };
   const recall = isLog ? `Log file \u2014 use Read with offset/limit params, or: token-goat bash-output --file "${filePath}" --tail 100 --grep "error|ERROR"` : "Use Read with offset and limit params to sample specific line ranges.";
   if (previewUnavailable(content, length)) {
     return {
@@ -18868,6 +18937,64 @@ ${fenceUntrustedFileContent(preview)}
 ${recall}`
   };
 }
+function handleSvg(filePath, content, contentLengthHint) {
+  const length = contentLengthHint ?? content.length;
+  if (length < FILE_TYPE_THRESHOLDS.svg) return { shouldBlock: false, message: "" };
+  if (previewUnavailable(content, length)) {
+    return {
+      shouldBlock: true,
+      message: [
+        `Large SVG file (${formatBytes(length)}) \u2014 too large to preview (exceeds the in-hook scan cap).`,
+        `Inspect structure: token-goat xml-outline "${filePath}"`,
+        `Query layers/elements: token-goat xml-query "${filePath}" "//g[@id]"`
+      ].join("\n")
+    };
+  }
+  const titleMatch = content.match(/<title[^>]*>([^<]*)<\/title>/i);
+  const title = titleMatch ? titleMatch[1]?.trim() : "";
+  const groupIds = [];
+  const idRegex = /<g[^>]*\bid=["']([^"']+)["']/gi;
+  let match;
+  while ((match = idRegex.exec(content)) !== null && groupIds.length < 10) {
+    if (match[1]) groupIds.push(match[1]);
+  }
+  const preview = [
+    title ? `Title: ${title}` : "",
+    groupIds.length > 0 ? `Layer/Group IDs: ${groupIds.join(", ")}` : ""
+  ].filter(Boolean).join("\n");
+  return {
+    shouldBlock: true,
+    message: [
+      `Large SVG/diagram file (${formatBytes(length)}) \u2014 raw coordinate paths flood context.`,
+      preview ? fenceUntrustedFileContent(preview) : "",
+      `Inspect structure: token-goat xml-outline "${filePath}"`,
+      `Query elements: token-goat xml-query "${filePath}" "//g[@id]"`,
+      `Search text labels: token-goat grep "<text>" --glob "${filePath}"`
+    ].filter(Boolean).join("\n")
+  };
+}
+function handleXml(filePath, content, contentLengthHint) {
+  const length = contentLengthHint ?? content.length;
+  if (length < FILE_TYPE_THRESHOLDS.xml) return { shouldBlock: false, message: "" };
+  if (previewUnavailable(content, length)) {
+    return {
+      shouldBlock: true,
+      message: [
+        `Large XML file (${formatBytes(length)}) \u2014 too large to preview (exceeds the in-hook scan cap).`,
+        `Inspect structure: token-goat xml-outline "${filePath}"`,
+        `Query nodes: token-goat xml-query "${filePath}" "<selector>"`
+      ].join("\n")
+    };
+  }
+  return {
+    shouldBlock: true,
+    message: [
+      `Large XML file (${formatBytes(length)}).`,
+      `Inspect hierarchy: token-goat xml-outline "${filePath}"`,
+      `Query specific elements: token-goat xml-query "${filePath}" "<selector>"`
+    ].join("\n")
+  };
+}
 function handleOfficeBinary(filePath) {
   const filename = filePath.split(/[\\/]/).pop() || "";
   const parts = filename.split(".");
@@ -18986,6 +19113,8 @@ function dispatchFileTypeHandler(filePath, content, contentLengthHint) {
   if (["md", "mdx", "markdown", "rst"].includes(ext)) return null;
   const effectiveLength = contentLengthHint ?? content.length;
   if (ext === "pdf") return handlePdf(filePath, effectiveLength);
+  if (ext === "svg") return handleSvg(filePath, content, effectiveLength);
+  if (ext === "xml") return handleXml(filePath, content, effectiveLength);
   if (["html", "htm", "xhtml"].includes(ext)) return handleHtml(filePath, content, effectiveLength);
   if (["txt", "log", "out", "err", "trace"].includes(ext)) return handleTxt(filePath, content, effectiveLength);
   if (ext === "xlsx") return handleXlsx(filePath);
@@ -19335,7 +19464,7 @@ function buildDeltaCapsule(projectRoot, limit = 8) {
   }).slice(0, limit);
   if (changed.length === 0) return null;
   return `Cross-session evidence changed since it was cached:
-${changed.map((entry) => `- ${entry.source} (use a fresh surgical read)`).join("\n")}`;
+${changed.map((entry) => `- ${displaySafePath(entry.source)} (use a fresh surgical read)`).join("\n")}`;
 }
 
 // src/notebook_compact.ts
@@ -19941,6 +20070,41 @@ function readIntToolInput(event, key) {
   }
   return void 0;
 }
+function readRequestedSliceWindow(event) {
+  const rawOffset = readIntToolInput(event, "offset");
+  const rawLimit = readIntToolInput(event, "limit");
+  if (rawOffset !== void 0 || rawLimit !== void 0) {
+    return {
+      offset: rawOffset !== void 0 && rawOffset >= 1 ? Math.floor(rawOffset) : 1,
+      limit: rawLimit !== void 0 && rawLimit > 0 ? Math.floor(rawLimit) : void 0,
+      isExplicitSlice: true
+    };
+  }
+  const rawRange = event.toolInput["view_range"] ?? event.toolInput["lines"] ?? event.toolInput["range"];
+  if (Array.isArray(rawRange) && rawRange.length >= 1) {
+    const rawStart = Number(rawRange[0]);
+    const startVal = Number.isFinite(rawStart) && rawStart >= 1 ? Math.floor(rawStart) : 1;
+    if (rawRange.length >= 2) {
+      const rawEnd = Number(rawRange[1]);
+      if (rawEnd === -1) {
+        return { offset: startVal, isExplicitSlice: true };
+      }
+      if (Number.isFinite(rawEnd) && rawEnd >= startVal) {
+        const lineCount = Math.floor(rawEnd - startVal + 1);
+        return { offset: startVal, limit: lineCount, isExplicitSlice: true };
+      }
+    }
+    return { offset: startVal, isExplicitSlice: true };
+  }
+  const startLine = readIntToolInput(event, "start_line") ?? readIntToolInput(event, "startLine") ?? readIntToolInput(event, "start");
+  const endLine = readIntToolInput(event, "end_line") ?? readIntToolInput(event, "endLine") ?? readIntToolInput(event, "end");
+  if (startLine !== void 0 || endLine !== void 0) {
+    const effectiveStart = startLine !== void 0 && startLine >= 1 ? Math.floor(startLine) : 1;
+    const effectiveLimit = endLine !== void 0 && endLine >= effectiveStart ? Math.floor(endLine - effectiveStart + 1) : void 0;
+    return { offset: effectiveStart, limit: effectiveLimit, isExplicitSlice: true };
+  }
+  return { isExplicitSlice: false };
+}
 function readStartLine(event) {
   const resp = event.raw["tool_response"];
   if (resp === null || typeof resp !== "object") return 1;
@@ -20004,11 +20168,10 @@ function scanRequestedSlice(absPath, offset, limit) {
   }
 }
 function estimateRequestedSlice(event, absPath) {
-  const offset = readIntToolInput(event, "offset");
-  const limit = readIntToolInput(event, "limit");
-  if (limit === void 0 || limit <= 0) return { kind: "unbounded" };
-  const effectiveOffset = offset !== void 0 && offset >= 1 ? offset : 1;
-  const scan = scanRequestedSlice(absPath, effectiveOffset, limit);
+  const window2 = readRequestedSliceWindow(event);
+  if (window2.limit === void 0 || window2.limit <= 0) return { kind: "unbounded" };
+  const effectiveOffset = window2.offset !== void 0 && window2.offset >= 1 ? window2.offset : 1;
+  const scan = scanRequestedSlice(absPath, effectiveOffset, window2.limit);
   if (scan === null) return { kind: "unbounded" };
   if (scan.nearSingleLine) return { kind: "nearSingleLine" };
   if (scan.trustworthy) return { kind: "bytes", bytes: scan.bytes };
@@ -20032,7 +20195,7 @@ function isSourceExtension(basename12) {
   return language === "apex" || language === "salesforce_metadata" || language === "salesforce_markup";
 }
 var BINARY_FILE_TYPE_EXTS = /* @__PURE__ */ new Set(["pdf", "docx", "xlsx", "pptx", "odt", "ods", "ott", "odp"]);
-var TEXT_FILE_TYPE_EXTS = /* @__PURE__ */ new Set(["html", "htm", "xhtml", "txt", "log", "out", "err", "trace", "csv", "tsv", "vtt", "srt"]);
+var TEXT_FILE_TYPE_EXTS = /* @__PURE__ */ new Set(["html", "htm", "xhtml", "txt", "log", "out", "err", "trace", "csv", "tsv", "vtt", "srt", "svg", "xml"]);
 var DISPATCHED_FILE_TYPE_EXTS = /* @__PURE__ */ new Set([...BINARY_FILE_TYPE_EXTS, ...TEXT_FILE_TYPE_EXTS]);
 function isDispatchedFileType(basename12) {
   return DISPATCHED_FILE_TYPE_EXTS.has(path21.extname(basename12).slice(1).toLowerCase());
@@ -20051,7 +20214,7 @@ function surgicalHint(filePath, basename12, lineCount) {
 }
 function detectSkillFile(filePath) {
   const match = filePath.match(/\.claude[\\/]skills[\\/]([^\\/]+)[\\/]SKILL\.md$/i);
-  return match ? match[1] : null;
+  return match ? displaySafePath(match[1]) : null;
 }
 function buildLineDiffDetailed(oldContent, newContent, label) {
   const oldLines = oldContent.split("\n");
@@ -20196,7 +20359,7 @@ function preReadHandlerInner(event) {
     } catch {
     }
   }
-  const basename12 = path21.basename(normalized);
+  const basename12 = displaySafePath(path21.basename(normalized));
   if (isLockFile(basename12)) {
     return denyOutput(
       'Lock files are rarely useful to read in full. Use `token-goat section "' + shown + '::<section>"` to extract a specific dependency, or read the relevant manifest instead.'
@@ -20248,7 +20411,7 @@ function preReadHandlerInner(event) {
     } catch {
     }
   }
-  const compactUnrangedRead = readIntToolInput(event, "offset") === void 0 && readIntToolInput(event, "limit") === void 0;
+  const compactUnrangedRead = !readRequestedSliceWindow(event).isExplicitSlice;
   if (event.toolName !== "Grep" && compactUnrangedRead && loadConfig().hints.stable_doc_compacts && _isDocFile(normalized)) {
     const compactPath = compactPathFor(normalized);
     if (isCompactFresh(compactPath, normalized)) {
@@ -20257,7 +20420,7 @@ function preReadHandlerInner(event) {
         recordActualRead(event, normalized);
         const fullSize = statSize(normalized) ?? 0;
         const savedBytes = counterfactualCredit(fullSize, compactBody.length);
-        recordStat("session_hint", savedBytes, savedTokensFromBytes(savedBytes));
+        recordStat("session_hint", savedBytes, savedTokensFromBytes(savedBytes), void 0, "stable-doc-compact");
         return denyOutput(
           "Serving the extractive compact sidecar in place of the full file (source unchanged since the last `compact-doc` build):\n\n" + fenceUntrustedFileContent(compactBody) + '\n\nThis is a lossy extract, not the whole file: front matter and any prose before the first heading are dropped entirely, each section is cut to its opening sentences, and long code fences are truncated. If you need content it left out, re-read with offset/limit for a line window, or `token-goat section "' + shown + '::Heading"` for one section in full. Use `token-goat compact-doc "' + shown + '" --force` to rebuild it, or `token-goat compact-doc "' + shown + '" --show` to view it directly. ' + editAnywayHint(normalized)
         );
@@ -20274,7 +20437,7 @@ function preReadHandlerInner(event) {
       if (savedBytes >= NB_STRIP_MIN_SAVINGS) {
         recordActualRead(event, normalized);
         const nbCredit = counterfactualCredit(rawBytes.length, sidecarContent.length);
-        recordStat("session_hint", nbCredit, savedTokensFromBytes(nbCredit));
+        recordStat("session_hint", nbCredit, savedTokensFromBytes(nbCredit), void 0, "notebook-strip");
         return denyOutput(
           "Serving the output-stripped notebook in place of the full file (code-cell outputs and execution counts removed; source and metadata preserved):\n\n" + fenceUntrustedFileContent(sidecarContent) + "\n\n" + editAnywayHint(normalized)
         );
@@ -20331,7 +20494,7 @@ function preReadHandlerInner(event) {
           }
           return denyOutput(message);
         }
-        const unrangedRead = readIntToolInput(event, "offset") === void 0 && readIntToolInput(event, "limit") === void 0;
+        const unrangedRead = !readRequestedSliceWindow(event).isExplicitSlice;
         if (loadConfig().hints.subagent_markdown_first_read_deny && event.agentId !== void 0 && unrangedRead && SUBAGENT_MD_FIRST_READ_DENY_EXT_RE.test(basename12) && markdownSize !== null && markdownSize >= SUBAGENT_MD_FIRST_READ_DENY_BYTES) {
           recordStat("subagent_markdown_first_read_deny", 0, 0);
           return denyOutput(
@@ -20402,7 +20565,7 @@ function preReadHandlerInner(event) {
       if (snapDiff.kind === "diff") {
         recordActualRead(event, normalized);
         const artifactDiffCredit = counterfactualCredit(snapDiff.currentContent.length, snapDiff.diff.length);
-        recordStat("session_hint", artifactDiffCredit, savedTokensFromBytes(artifactDiffCredit));
+        recordStat("session_hint", artifactDiffCredit, savedTokensFromBytes(artifactDiffCredit), void 0, "artifact-snapshot-diff");
         return denyOutput(
           "Content changed since last read of " + basename12 + ". Here is what changed:\n\n" + fenceUntrustedFileContent("```diff\n" + snapDiff.diff + "\n```") + "\n\n" + sessionArtifactRecall(normalized)
         );
@@ -20420,7 +20583,7 @@ function preReadHandlerInner(event) {
       recordActualRead(event, normalized);
       if (outputSize !== null && outputSize >= TASK_OUTPUT_DENY_BYTES) {
         const artifactDenyCredit = counterfactualCredit(outputSize);
-        recordStat("session_hint", artifactDenyCredit, savedTokensFromBytes(artifactDenyCredit));
+        recordStat("session_hint", artifactDenyCredit, savedTokensFromBytes(artifactDenyCredit), void 0, "artifact-large-deny");
         return denyOutput(
           label + " is large (" + toKB(outputSize) + "KB). " + sessionArtifactRecall(normalized)
         );
@@ -20509,12 +20672,12 @@ function preReadHandlerInner(event) {
     if (config2.hints.reread_deny && !protectedRead) {
       if (wasFileTruncatedThisSession(normalized)) {
         if (estimateTruncatedLineCount(normalized) >= config2.hints.truncated_read_min_lines) {
-          recordStat("session_hint", rereadCredit, savedTokensFromBytes(rereadCredit));
+          recordStat("session_hint", rereadCredit, savedTokensFromBytes(rereadCredit), void 0, "reread-truncated-deny");
           return denyOutput(truncatedReadDenyMessage(normalized));
         }
       }
       if (/\.(md|mdx|markdown|rst)$/i.test(basename12)) {
-        recordStat("session_hint", rereadCredit, savedTokensFromBytes(rereadCredit));
+        recordStat("session_hint", rereadCredit, savedTokensFromBytes(rereadCredit), void 0, "reread-doc-deny");
         return denyOutput(
           'Markdown file already read this session. Use `token-goat section "' + shown + '::HeadingName"` to read one section.'
         );
@@ -20530,7 +20693,7 @@ function preReadHandlerInner(event) {
     }
     const hint = _isDocFile(normalized) ? 'Use `token-goat section "' + shown + '::SectionName"` to read one section.' : "Use token-goat read/section/symbol to re-read surgically.";
     if (config2.hints.reread_deny && !protectedRead && (rereadBytes >= config2.hints.reread_deny_min_bytes || reads >= 2)) {
-      recordStat("session_hint", rereadCredit, savedTokensFromBytes(rereadCredit));
+      recordStat("session_hint", rereadCredit, savedTokensFromBytes(rereadCredit), void 0, "reread-count-deny");
       return denyOutput(
         shown + " was already read this session (" + reads + " " + plural + "). " + hint
       );
@@ -20555,7 +20718,7 @@ function preReadHandlerInner(event) {
     const hint = _isDocFile(normalized) ? 'Use `token-goat section "' + shown + '::SectionName"` to read one section.' : "Consider token-goat skeleton or token-goat section.";
     if (gateSize >= largeFileDenyBytes()) {
       const denyCredit = counterfactualCredit(size);
-      recordStat("session_hint", denyCredit, savedTokensFromBytes(denyCredit));
+      recordStat("session_hint", denyCredit, savedTokensFromBytes(denyCredit), void 0, "large-file-deny");
       return denyOutput(
         shown + " is very large (" + kb + "KB). " + hint + " " + describeSliceAdvice(slice, normalized) + " " + editAnywayHint(normalized)
       );
@@ -20701,9 +20864,9 @@ function readWindowFromDisk(event, normalized) {
   const size = statSize(normalized);
   if (size === null || size > SLICE_ESTIMATE_SCAN_CAP_BYTES) return null;
   const text = decodeSource(fs21.readFileSync(normalized));
-  const limit = readIntToolInput(event, "limit");
-  const offset = readIntToolInput(event, "offset");
-  const start = offset !== void 0 && offset >= 1 ? offset : 1;
+  const window2 = readRequestedSliceWindow(event);
+  const start = window2.offset !== void 0 && window2.offset >= 1 ? window2.offset : 1;
+  const limit = window2.limit;
   if (start === 1 && (limit === void 0 || limit <= 0)) return text;
   const lines2 = text.split("\n");
   const from = start - 1;
@@ -20826,7 +20989,7 @@ function elideAlreadyServedLines(event, respText) {
 function structuralFoldInputs(event, respText) {
   const filePath = getFilePath(event);
   if (filePath === void 0) return null;
-  if (readIntToolInput(event, "offset") !== void 0 || readIntToolInput(event, "limit") !== void 0) return null;
+  if (readRequestedSliceWindow(event).isExplicitSlice) return null;
   if (isTruncatedReadDelivery(event, respText)) return null;
   if (redactSecrets(respText).count > 0) return null;
   const parsed = parseNumberedReadResult(respText, readStartLine(event));
@@ -20861,14 +21024,15 @@ function foldCodeBodies(event, respText) {
   if (filePath === void 0) return null;
   const normalized = normalizePath(filePath);
   if (isImagePath(normalized)) return null;
-  const requestedOffset = readIntToolInput(event, "offset");
+  const sliceWin = readRequestedSliceWindow(event);
+  const requestedOffset = sliceWin.offset;
   if (requestedOffset !== void 0 && readStartLine(event) !== requestedOffset) return null;
   if (isTruncatedReadDelivery(event, respText)) return null;
   if (redactSecrets(respText).count > 0) return null;
   const parsed = parseNumberedReadResult(respText, readStartLine(event));
   if (parsed === null) return null;
   const shown = displaySafePath(toDisplayPath(findProject(getCwd(event) ?? process.cwd())?.root, normalized));
-  const folded = foldDelivery(parsed.rows, normalized, shown, requestedOffset !== void 0 || readIntToolInput(event, "limit") !== void 0);
+  const folded = foldDelivery(parsed.rows, normalized, shown, sliceWin.isExplicitSlice);
   if (folded === null) return null;
   const rewritten = [...parsed.header, fenceUntrustedFileContent(folded.numbered.join("\n")), ...parsed.trailer].join("\n");
   const originalBytes = Buffer.byteLength(respText, "utf-8");
@@ -21991,6 +22155,7 @@ export {
   clearCurlDownload,
   recordFileLineRange,
   getFileLineRanges,
+  GENERIC_SERVED_OUTPUT_KEY,
   recordFileServedOutput,
   getFileServedOutputs,
   markFileTruncated,
@@ -22099,6 +22264,9 @@ export {
   PARSER_FINGERPRINT,
   stripLeadingAttributes2 as stripLeadingAttributes,
   IMPORT_RE2 as IMPORT_RE,
+  treeSitterCoreAvailable,
+  treeSitterCoreLoadError,
+  isTreeSitterAvailable,
   yamlOpenQuoteAfter,
   yamlLineClosesQuote,
   lineOpenDelimiterAfter,

package/dist/{token-goat-chunk-3ZAMYHQG.mjs → token-goat-chunk-DRH4CVGF.mjs}

@@ -7,11 +7,11 @@ import {
   expandGlobs,
   leftoverIntegrations,
   run
-} from "./token-goat-chunk-J7LGMKF3.mjs";
-import "./token-goat-chunk-46VKOCUH.mjs";
-import "./token-goat-chunk-7GJBID7S.mjs";
-import "./token-goat-chunk-3HJQR4OO.mjs";
-import "./token-goat-chunk-T2OE7MYM.mjs";
+} from "./token-goat-chunk-PNMGVJ4C.mjs";
+import "./token-goat-chunk-L6Q2XM5X.mjs";
+import "./token-goat-chunk-QCHD2ENV.mjs";
+import "./token-goat-chunk-BYDDWQ2T.mjs";
+import "./token-goat-chunk-KIEOFWLL.mjs";
 import "./token-goat-chunk-EEIDFMEM.mjs";
 import "./token-goat-chunk-A37V4PBF.mjs";
 export {

package/dist/{token-goat-chunk-TX64YIIV.mjs → token-goat-chunk-K5GKX6ND.mjs}

@@ -4,14 +4,14 @@ import {
   buildEvent,
   relay,
   relayInProcess
-} from "./token-goat-chunk-EIZYCVBM.mjs";
+} from "./token-goat-chunk-P5SQX2VK.mjs";
 import {
   MAX_STDIN_BYTES,
   readStdinJson
-} from "./token-goat-chunk-7GJBID7S.mjs";
-import "./token-goat-chunk-3HJQR4OO.mjs";
-import "./token-goat-chunk-LXIC7MTW.mjs";
-import "./token-goat-chunk-T2OE7MYM.mjs";
+} from "./token-goat-chunk-QCHD2ENV.mjs";
+import "./token-goat-chunk-BYDDWQ2T.mjs";
+import "./token-goat-chunk-OBDTBOQA.mjs";
+import "./token-goat-chunk-KIEOFWLL.mjs";
 import "./token-goat-chunk-EEIDFMEM.mjs";
 import "./token-goat-chunk-A37V4PBF.mjs";
 export {