@oh-my-pi/pi-coding-agent 13.16.3 → 13.16.4

package/CHANGELOG.md

@@ -2,6 +2,17 @@
 
 ## [Unreleased]
 
+## [13.16.4] - 2026-03-28
+### Changed
+
+- Renamed hashline helper functions from `hlineref`/`hlinefull` to `href`/`hline` for brevity
+- Simplified hashline edit location API: replaced separate `line` and `block` properties with unified `range` property accepting `{ pos, end }` for all range-based edits
+- Updated hashline prompt documentation to reflect new `range` syntax and clarified editing guidelines
+
+### Fixed
+
+- Added detection for `kysely-codegen` generated files in auto-generated file guard
+
 ## [13.16.1] - 2026-03-27
 
 ### Added
@@ -17,6 +28,7 @@
 - Updated grep tool description to remove ripgrep-specific implementation detail
 
 ## [13.16.0] - 2026-03-27
+
 ### Added
 
 - Implemented root path alias: bare `/` in tool inputs now resolves to the session working directory instead of the filesystem root
@@ -36,6 +48,7 @@
 - Added configurable `app.model.selectTemporary` keybinding for temporary model selection.
 
 ## [13.15.0] - 2026-03-23
+
 ### Breaking Changes
 
 - Changed hashline edit schema from flat `op`/`pos`/`end`/`lines` fields to structured `loc`/`content` format with location-specific objects
@@ -162,10 +175,10 @@
 - Fixed autoresearch logging to require durable ASI metadata (hypothesis, rollback_reason, next_action_hint) for every run including rollback context for discarded, crashed, and checks-failed experiments
 - Fixed autoresearch logging to require durable ASI metadata for every run, including rollback context for discarded, crashed, and checks-failed experiments
 
-
 ### Fixed
 
 - Fixed resumed and session-switched GitHub Copilot/OpenAI Responses conversations replaying stale assistant native history from older saved sessions by sanitizing persisted assistant replay metadata on rehydration and resetting provider session state across live session boundaries ([#505](https://github.com/can1357/oh-my-pi/issues/505))
+
 ## [13.14.0] - 2026-03-20
 
 ### Added
@@ -200,6 +213,7 @@
 - Added `mcpServerName` and `mcpToolName` optional properties to custom tools for MCP server discovery and search metadata
 
 ## [13.13.1] - 2026-03-18
+
 ### Added
 
 - Automatic deduplication of identical context files by content, keeping the closest (lowest depth) copy when duplicates are discovered
@@ -218,6 +232,7 @@
 - Enhanced auto-generated marker detection to only scan leading header comments rather than entire file prefix, improving accuracy for files with generated markers in code
 
 ## [13.12.10] - 2026-03-17
+
 ### Added
 
 - Added `args` field to ShellResult to capture the executed command
@@ -230,6 +245,7 @@
 - Updated `run()` command documentation to clarify available ShellResult fields
 
 ## [13.12.9] - 2026-03-17
+
 ### Added
 
 - Added `/session delete` command to delete current session with confirmation and return to session selector
@@ -274,6 +290,7 @@
 - Fixed session directory resolution to correctly handle symlink-equivalent paths, ensuring aliased home and temp directories resolve to the same session storage location as their real targets
 
 ## [13.12.7] - 2026-03-16
+
 ### Changed
 
 - Modified `getSelectedMCPToolNames()` to return only active MCP tools in non-discovery sessions, filtering by tool registry availability
@@ -285,6 +302,7 @@
 - Fixed MCP tool selection tracking to properly distinguish between discovery-enabled and non-discovery sessions, preventing orphaned tool selections after manual deactivation
 
 ## [13.12.6] - 2026-03-15
+
 ### Changed
 
 - Updated llama.cpp model discovery to read context window from the `/props` endpoint's `default_generation_settings.n_ctx` field instead of using hardcoded 128000 default
@@ -320,6 +338,7 @@
 - Fixed automatic migration of legacy session directories to new `-tmp-` prefixed naming scheme for temp-root sessions
 
 ## [13.12.4] - 2026-03-15
+
 ### Added
 
 - Exposed `settings` instance in `CustomToolContext` for session-specific configuration access
@@ -357,6 +376,7 @@
 - Fixed skill loading to properly respect disabled skill names when loading from custom directories
 
 ## [13.12.1] - 2026-03-15
+
 ### Added
 
 - Support for move-only operations that preserve exact bytes including binary files
@@ -396,7 +416,7 @@
 - Moved LSP settings (lsp.enabled, lsp.formatOnWrite, lsp.diagnosticsOnWrite, lsp.diagnosticsOnEdit) to the Editing tab
 - Moved bash interceptor settings to the Editing tab
 - Moved Python settings (python.toolMode, python.kernelMode, python.sharedGateway) to the Editing tab
-- Moved task delegation settings (task.isolation.*, task.eager, task.maxConcurrency, task.maxRecursionDepth) to the Tasks tab
+- Moved task delegation settings (task.isolation.\*, task.eager, task.maxConcurrency, task.maxRecursionDepth) to the Tasks tab
 - Moved skill and command settings to the Tasks tab
 - Moved provider selection settings (providers.webSearch, providers.codeSearch, providers.image, etc.) to the Providers tab
 - Moved Exa settings to the Providers tab
@@ -452,6 +472,7 @@
 - Fixed line number parsing in compact diff preview to handle variable-width line number fields with leading whitespace
 
 ## [13.11.0] - 2026-03-12
+
 ### Added
 
 - Added Parallel as a web search provider with support for fast and research modes
@@ -494,6 +515,7 @@
 - Per-role `modelRoles` thinking selectors now propagate through commit/title helper model selection, legacy commit analysis, and agentic commit sessions while preserving default thinking inheritance when no role override is configured
 
 ## [13.10.1] - 2026-03-10
+
 ### Added
 
 - Exported `submitInteractiveInput()` function for programmatic submission of user input in interactive mode
@@ -501,15 +523,19 @@
 - Added reactive 401/403 retry with automatic token refresh on HTTP MCP transports
 - Added `refreshMCPOAuthToken()` for standard OAuth 2.0 refresh_token grants
 - Persisted `tokenUrl`, `clientId`, and `clientSecret` in MCP auth config for cross-session token refresh
+
 ### Fixed
+
 - Respected `PI_CONFIG_DIR` when discovering native user config paths for slash commands and related config directories ([#349](https://github.com/can1357/oh-my-pi/issues/349))
 
 ## [13.10.0] - 2026-03-10
+
 ### Fixed
 
 - Preserved text signature metadata (id and phase) when building OpenAI native history during session compaction
 
 ## [13.9.16] - 2026-03-10
+
 ### Breaking Changes
 
 - Web search tool no longer accepts `provider` parameter in tool calls; use internal provider resolution instead
@@ -548,6 +574,7 @@
 - Fixed model selector to show discovery status messages when provider has no models
 
 ## [13.9.15] - 2026-03-10
+
 ### Added
 
 - Added `ensureLoadingAnimation()` method to manage loading animation lifecycle and prevent duplicate spinners
@@ -558,6 +585,7 @@
 - Updated `showError()` to properly clean up loading animation state when errors occur
 
 ## [13.9.12] - 2026-03-09
+
 ### Added
 
 - Added Tavily as a supported web search provider with `TAVILY_API_KEY` credential discovery and provider fallback support
@@ -575,6 +603,7 @@
 - Canonicalized bash executor working directories before handing them to brush so `pwd` stays aligned with canonical Git worktree paths in symlinked workspaces
 
 ## [13.9.10] - 2026-03-08
+
 ### Added
 
 - Added `env` parameter to bash tool to pass environment variables safely without shell re-parsing, preventing quote and special character bugs with multiline or untrusted values
@@ -586,6 +615,7 @@
 - Changed bash tool to display environment variable assignments in command preview when `env` parameter is used
 
 ## [13.9.8] - 2026-03-08
+
 ### Added
 
 - Added docs.rs scraper for extracting Rust crate documentation from rustdoc JSON, including support for modules, functions, structs, traits, enums, and other Rust items with caching
@@ -620,6 +650,7 @@
 - Updated `grep` tool to combine glob patterns from `path` and `glob` parameters instead of throwing an error when both are provided
 
 ## [13.9.4] - 2026-03-07
+
 ### Added
 
 - Automatic detection of Ollama model capabilities including reasoning/thinking support and vision input via the `/api/show` endpoint
@@ -708,6 +739,7 @@
 - Auto-corrected off-by-one range start errors in hashline edits that would duplicate preceding lines
 
 ## [13.9.0] - 2026-03-05
+
 ### Added
 
 - Added `read.defaultLimit` setting to configure default number of lines returned by read tool when no limit is specified (default: 300 lines)
@@ -736,6 +768,7 @@
 - Fixed provider session state not being cleared when branching or navigating tree history, preventing resource leaks with codex provider sessions
 
 ## [13.8.0] - 2026-03-04
+
 ### Added
 
 - Added `buildCompactHashlineDiffPreview()` function to generate compact diff previews for model-visible tool responses, collapsing long unchanged runs and consecutive additions/removals to show edit shape without full file content
@@ -751,6 +784,7 @@
 - Fixed `:thinking` suffix in `modelRoles` config values silently breaking model resolution (e.g., `slow: anthropic/claude-opus-4-6:high`) and being stripped on Ctrl+P role cycling
 
 ## [13.7.6] - 2026-03-04
+
 ### Added
 
 - Exported `dedupeParseErrors` utility function to deduplicate parse error messages while preserving order
@@ -761,7 +795,9 @@
 - Normalized parse error output in ast-grep to remove pattern-specific prefixes and show only file-level errors
 
 ## [13.7.4] - 2026-03-04
+
 ### Added
+
 - Added `fetch.useKagiSummarizer` setting to toggle Kagi Universal Summarizer usage in the fetch tool.
 
 ### Fixed
@@ -781,7 +817,7 @@
 ### Changed
 
 - Updated hashline prompt documentation with clearer operation syntax and improved examples showing full edit structure with path and edits array
-- Refactored `hlineref` Handlebars helper to return JSON-quoted strings for safer embedding in JSON blocks within prompts
+- Refactored `href` Handlebars helper to return JSON-quoted strings for safer embedding in JSON blocks within prompts
 - Improved `hashlineParseText` to correctly preserve blank lines and trailing empty strings in array input while stripping trailing newlines from string input
 - Optimized duplicate line detection in range replacements to use trimmed comparison, reducing false positives from whitespace differences
 - Refactored Kagi search provider to use shared Kagi API utilities from `web/kagi` module
@@ -792,6 +828,7 @@
 - Fixed `isEscapedTabAutocorrectEnabled` environment variable parsing to use switch statement for clearer logic and consistent default behavior
 
 ## [13.7.2] - 2026-03-04
+
 ### Added
 
 - Added support for direct OAuth provider login via `/login <provider>` command (e.g., `/login kagi`)
@@ -808,6 +845,7 @@
 - Fixed `ask` timeout handling to auto-select the recommended option instead of aborting the turn, while preserving explicit user-cancel abort behavior ([#266](https://github.com/can1357/oh-my-pi/issues/266))
 
 ## [13.6.2] - 2026-03-03
+
 ### Fixed
 
 - Fixed LM Studio API key retrieval to use configured provider name instead of hardcoded 'lm-studio'
@@ -822,7 +860,9 @@
 - Fixed `omp update` silently succeeding without actually updating the binary when the update channel (bun global vs compiled binary) doesn't match the installation method ([#247](https://github.com/can1357/oh-my-pi/issues/247))
 - Added post-update verification that checks the resolved `omp` binary reports the expected version, with actionable warnings on mismatch
 - `omp update` now detects when the `omp` in PATH is not managed by bun and falls back to binary replacement instead of updating the wrong location
+
 ## [13.6.0] - 2026-03-03
+
 ### Added
 
 - Added `mcp://` internal URL protocol for reading MCP server resources directly via the read tool (e.g., `read(path="mcp://resource-uri")`)
@@ -853,9 +893,11 @@
 - Fixed URI template matching to correctly handle expressions that expand to empty strings
 
 ## [13.5.6] - 2026-03-01
+
 ### Changed
 
 - Updated OAuth client name from 'oh-my-pi MCP' to 'Codex' for dynamic client registration
+
 ### Fixed
 
 - Fixed exit_plan_mode handler to abort active agent turn before opening plan approval selector, ensuring proper session cleanup
@@ -867,6 +909,7 @@
 - Added Kagi web search provider (Search API v0) with related searches support and automatic `KAGI_API_KEY` detection
 
 ## [13.5.4] - 2026-03-01
+
 ### Added
 
 - Added `authServerUrl` field to `AuthDetectionResult` to capture OAuth server metadata from `Mcp-Auth-Server` headers
@@ -875,6 +918,7 @@
 - Added recursive auth server discovery to follow `authorization_servers` references when discovering OAuth endpoints
 
 - Added `omp agents unpack` CLI subcommand to export bundled subagent definitions to `~/.omp/agent/agents` by default, with `--project` support for `./.omp/agents`
+
 ### Changed
 
 - Enhanced `discoverOAuthEndpoints()` to accept optional `authServerUrl` parameter and query both auth server and resource server for OAuth metadata
@@ -1569,7 +1613,7 @@
 - Updated hashline reference format from `LINE:HASH` to `LINE#ID` throughout the codebase for improved clarity
 - Renamed hashline edit operations: `set_line` → `set`, `replace_lines` → `set_range`, `insert_after` → `insert` with support for `before` and `between` anchors
 - Changed hashline edit `body` field from string to array of strings for clearer multiline handling
-- Updated handlebars helpers: renamed `hashline` to `hlineref` and added `hlinefull` for formatted line output
+- Updated handlebars helpers: renamed `hashline` to `href` and added `hline` for formatted line output
 - Improved insert operation to support `before`, `after`, and `between` (both anchors) positioning modes
 - Made autocorrect heuristics (boundary echo stripping, indent restoration) conditional on `PI_HL_AUTOCORRECT` environment variable
 - Updated SSH host discovery to load from managed omp config paths (.omp/ssh.json and ~/.omp/agent/ssh.json) in addition to legacy root-level ssh.json and .ssh.json files
@@ -1901,7 +1945,6 @@
 - Improved error reporting in fetch tool to include HTTP status codes when URL fetching fails
 - Fixed fetch tool to preserve actual response metadata (finalUrl, contentType) instead of defaults when requests fail
 
-
 ## [12.1.0] - 2026-02-13
 
 ### Added

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "13.16.3",
+	"version": "13.16.4",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://github.com/can1357/oh-my-pi",
 	"author": "Can Boluk",
@@ -42,12 +42,12 @@
 	"dependencies": {
 		"@agentclientprotocol/sdk": "0.16.1",
 		"@mozilla/readability": "^0.6",
-		"@oh-my-pi/omp-stats": "13.16.3",
-		"@oh-my-pi/pi-agent-core": "13.16.3",
-		"@oh-my-pi/pi-ai": "13.16.3",
-		"@oh-my-pi/pi-natives": "13.16.3",
-		"@oh-my-pi/pi-tui": "13.16.3",
-		"@oh-my-pi/pi-utils": "13.16.3",
+		"@oh-my-pi/omp-stats": "13.16.4",
+		"@oh-my-pi/pi-agent-core": "13.16.4",
+		"@oh-my-pi/pi-ai": "13.16.4",
+		"@oh-my-pi/pi-natives": "13.16.4",
+		"@oh-my-pi/pi-tui": "13.16.4",
+		"@oh-my-pi/pi-utils": "13.16.4",
 		"@sinclair/typebox": "^0.34",
 		"@xterm/headless": "^6.0",
 		"ajv": "^8.18",

package/src/config/prompt-templates.ts

@@ -258,19 +258,19 @@ function formatHashlineRef(lineNum: unknown, content: unknown): { num: number; t
 }
 
 /**
- * {{hlineref lineNum "content"}} — compute a real hashline ref for prompt examples.
+ * {{href lineNum "content"}} — compute a real hashline ref for prompt examples.
  * Returns `"lineNum#hash"` using the actual hash algorithm.
  */
-handlebars.registerHelper("hlineref", (lineNum: unknown, content: unknown): string => {
+handlebars.registerHelper("href", (lineNum: unknown, content: unknown): string => {
 	const { ref } = formatHashlineRef(lineNum, content);
 	return JSON.stringify(ref);
 });
 
 /**
- * {{hlinefull lineNum "content"}} — format a full read-style line with prefix.
+ * {{hline lineNum "content"}} — format a full read-style line with prefix.
  * Returns `"lineNum#hash:content"`.
  */
-handlebars.registerHelper("hlinefull", (lineNum: unknown, content: unknown): string => {
+handlebars.registerHelper("hline", (lineNum: unknown, content: unknown): string => {
 	const { ref, text } = formatHashlineRef(lineNum, content);
 	return `${ref}:${text}`;
 });

package/src/patch/index.ts

@@ -187,12 +187,9 @@ const locSchema = Type.Union(
 		Type.Object({ append: Type.String({ description: "anchor" }) }),
 		Type.Object({ prepend: Type.String({ description: "anchor" }) }),
 		Type.Object({
-			line: Type.String({ description: "anchor" }),
-		}),
-		Type.Object({
-			block: Type.Object({
-				pos: Type.String({ description: "anchor" }),
-				end: Type.String({ description: "limit position" }),
+			range: Type.Object({
+				pos: Type.String({ description: "first line to edit (inclusive)" }),
+				end: Type.String({ description: "last line to edit (inclusive)" }),
 			}),
 		}),
 	],
@@ -231,8 +228,7 @@ export type HashlineParams = Static<typeof hashlineEditParamsSchema>;
  * loc can be:
  *   - "append" / "prepend" — file-level insert
  *   - { append: anchor } / { prepend: anchor } — insert relative to anchor
- *   - { replace_line: anchor } — replace one line
- *   - { replace_block: { pos, end } } — replace inclusive range
+ *   - { range: { pos, end } } — replace inclusive range
  */
 function resolveEditAnchors(edits: HashlineToolEdit[]): HashlineEdit[] {
 	const result: HashlineEdit[] = [];
@@ -253,17 +249,13 @@ function resolveEditAnchors(edits: HashlineToolEdit[]): HashlineEdit[] {
 				const anchor = tryParseTag(loc.prepend);
 				if (!anchor) throw new Error("prepend requires a valid anchor.");
 				result.push({ op: "prepend_at", pos: anchor, lines });
-			} else if ("line" in loc) {
-				const anchor = tryParseTag(loc.line);
-				if (!anchor) throw new Error("line requires a valid anchor.");
-				result.push({ op: "replace_line", pos: anchor, lines });
-			} else if ("block" in loc) {
-				const posAnchor = tryParseTag(loc.block.pos);
-				const endAnchor = tryParseTag(loc.block.end);
-				if (!posAnchor || !endAnchor) throw new Error("block requires valid pos and end anchors.");
+			} else if ("range" in loc) {
+				const posAnchor = tryParseTag(loc.range.pos);
+				const endAnchor = tryParseTag(loc.range.end);
+				if (!posAnchor || !endAnchor) throw new Error("range requires valid pos and end anchors.");
 				result.push({ op: "replace_range", pos: posAnchor, end: endAnchor, lines });
 			} else {
-				throw new Error("Unknown loc shape. Expected append, prepend, line, or block.");
+				throw new Error("Unknown loc shape. Expected append, prepend, or range.");
 			}
 		} else {
 			throw new Error(`Invalid loc value: ${JSON.stringify(loc)}`);

package/src/patch/shared.ts

@@ -164,9 +164,8 @@ function formatStreamingHashlineEdits(edits: Partial<HashlineToolEdit>[], uiThem
 			return { srcLabel: `• ${loc} (file-level)`, dst: contentLines };
 		}
 		if (typeof loc === "object" && loc) {
-			if ("block" in loc && typeof loc.block === "object" && loc.block) {
-				const rb = loc.block as { pos?: string; end?: string };
-				return { srcLabel: `• block ${rb.pos ?? "?"}…${rb.end ?? "?"}`, dst: contentLines };
+			if ("range" in loc && typeof loc.range === "object" && loc.range) {
+				return { srcLabel: `• range ${loc.range.pos ?? "?"}…${loc.range.end ?? "?"}`, dst: contentLines };
 			}
 			if ("line" in loc) {
 				return { srcLabel: `• line ${(loc as { line: string }).line}`, dst: contentLines };

package/src/prompts/tools/hashline.md

@@ -16,40 +16,39 @@ Read the file first. Copy anchors exactly from the latest `read` output. In one
 **`loc` values**
 - `"append"` / `"prepend"` — insert at end/start of file
 - `{ append: "N#ID" }` / `{ prepend: "N#ID" }` — insert after/before anchored line
-- `{ line: "N#ID" }` — replace exactly one anchored line
-- `{ block: { pos: "N#ID", end: "N#ID" } }` — replace inclusive `pos..end`
-</operations>
+- `{ range: { pos: "N#ID", end: "N#ID" } }` — replace inclusive range of lines `pos..end` with new content
+  </operations>
 
 <examples>
-All examples below reference the same file, `util.ts`:
-```ts
-{{hlinefull  1 "// @ts-ignore"}}
-{{hlinefull  2 "const timeout = 5000;"}}
-{{hlinefull  3 "const tag = \"DO NOT SHIP\";"}}
-{{hlinefull  4 ""}}
-{{hlinefull  5 "function alpha() {"}}
-{{hlinefull  6 "\tlog();"}}
-{{hlinefull  7 "}"}}
-{{hlinefull  8 ""}}
-{{hlinefull  9 "function beta() {"}}
-{{hlinefull 10 "\t// TODO: remove after migration"}}
-{{hlinefull 11 "\tlegacy();"}}
-{{hlinefull 12 "\ttry {"}}
-{{hlinefull 13 "\t\treturn parse(data);"}}
-{{hlinefull 14 "\t} catch (err) {"}}
-{{hlinefull 15 "\t\tconsole.error(err);"}}
-{{hlinefull 16 "\t\treturn null;"}}
-{{hlinefull 17 "\t}"}}
-{{hlinefull 18 "}"}}
+All examples below reference the same file:
+```ts title="a.ts"
+{{hline  1 "// @ts-ignore"}}
+{{hline  2 "const timeout = 5000;"}}
+{{hline  3 "const tag = \"DO NOT SHIP\";"}}
+{{hline  4 ""}}
+{{hline  5 "function alpha() {"}}
+{{hline  6 "\tlog();"}}
+{{hline  7 "}"}}
+{{hline  8 ""}}
+{{hline  9 "function beta() {"}}
+{{hline 10 "\t// TODO: remove after migration"}}
+{{hline 11 "\tlegacy();"}}
+{{hline 12 "\ttry {"}}
+{{hline 13 "\t\treturn parse(data);"}}
+{{hline 14 "\t} catch (err) {"}}
+{{hline 15 "\t\tconsole.error(err);"}}
+{{hline 16 "\t\treturn null;"}}
+{{hline 17 "\t}"}}
+{{hline 18 "}"}}
 ```
 
 <example name="replace a block body">
 Replace only the catch body. Do not target the shared boundary line `} catch (err) {`.
 ```
 {
-  path: "util.ts",
+  path: "a.ts",
   edits: [{
-    loc: { block: { pos: {{hlineref 15 "\t\tconsole.error(err);"}}, end: {{hlineref 16 "\t\treturn null;"}} } },
+    loc: { range: { pos: {{href 15 "\t\tconsole.error(err);"}}, end: {{href 16 "\t\treturn null;"}} } },
     content: [
       "\t\tif (isEnoent(err)) return null;",
       "\t\tthrow err;"
@@ -60,12 +59,12 @@ Replace only the catch body. Do not target the shared boundary line `} catch (er
 </example>
 
 <example name="replace whole block including closing brace">
-Replace the entire body of `alpha`, including its closing `}`. `end` **MUST** be {{hlineref 7 "}"}} because `content` includes `}`.
+Replace the entire body of `alpha`, including its closing `}`. `end` **MUST** be {{href 7 "}"}} because `content` includes `}`.
 ```
 {
-  path: "util.ts",
+  path: "a.ts",
   edits: [{
-    loc: { block: { pos: {{hlineref 6 "\tlog();"}}, end: {{hlineref 7 "}"}} } },
+    loc: { range: { pos: {{href 6 "\tlog();"}}, end: {{href 7 "}"}} } },
     content: [
       "\tvalidate();",
       "\tlog();",
@@ -74,15 +73,15 @@ Replace the entire body of `alpha`, including its closing `}`. `end` **MUST** be
   }]
 }
 ```
-**Wrong**: using `end: {{hlineref 6 "\tlog();"}}` with the same content — line 7 (`}`) survives the replacement AND content emits `}`, producing two closing braces.
+**Wrong**: using `end: {{href 6 "\tlog();"}}` with the same content — line 7 (`}`) survives the replacement AND content emits `}`, producing two closing braces.
 </example>
 
 <example name="replace one line">
 ```
 {
-  path: "util.ts",
+  path: "a.ts",
   edits: [{
-    loc: { line: {{hlineref 2 "const timeout = 5000;"}} },
+    loc: { range: { pos: {{href 2 "const timeout = 5000;"}}, end: {{href 2 "const timeout = 5000;"}} } },
     content: ["const timeout = 30_000;"]
   }]
 }
@@ -92,9 +91,9 @@ Replace the entire body of `alpha`, including its closing `}`. `end` **MUST** be
 <example name="delete a range">
 ```
 {
-  path: "util.ts",
+  path: "a.ts",
   edits: [{
-    loc: { block: { pos: {{hlineref 10 "\t// TODO: remove after migration"}}, end: {{hlineref 11 "\tlegacy();"}} } },
+    loc: { range: { pos: {{href 10 "\t// TODO: remove after migration"}}, end: {{href 11 "\tlegacy();"}} } },
     content: null
   }]
 }
@@ -105,9 +104,9 @@ Replace the entire body of `alpha`, including its closing `}`. `end` **MUST** be
 When adding a sibling declaration, prefer `prepend` on the next declaration.
 ```
 {
-  path: "util.ts",
+  path: "a.ts",
   edits: [{
-    loc: { prepend: {{hlineref 9 "function beta() {"}} },
+    loc: { prepend: {{href 9 "function beta() {"}} },
     content: [
       "function gamma() {",
       "\tvalidate();",
@@ -123,10 +122,10 @@ When adding a sibling declaration, prefer `prepend` on the next declaration.
 <critical>
 - Make the minimum exact edit. Do not rewrite nearby code unless the consumed range requires it.
 - Use anchors exactly as `N#ID` from the latest `read` output.
-- `block` requires both `pos` and `end`. Other anchored ops require one anchor.
+- `range` requires both `pos` and `end`.
 - When your replacement `content` ends with a closing delimiter (`}`, `*/`, `)`, `]`), verify `end` includes the original line carrying that delimiter. If `end` stops one line too early, the original delimiter survives and your content adds a second copy.
 - **Self-check**: compare the last line of `content` with the line immediately after `end` in the file. If they match (e.g., both are `}`), extend `end` to include that line.
-- For a block, either replace only the body or replace the whole block. Do not split block boundaries.
+- For a range, either replace only the body or replace the whole range. Do not split range boundaries.
 - `content` must be literal file content with matching indentation. If the file uses tabs, use real tabs.
-- Do not use this tool to reformat or clean up unrelated code.
-</critical>
+- You **MUST NOT** use this tool to reformat or clean up unrelated code. **ALWAYS** use project-specific tooling like linters or code formatters which are much more efficient and reliable.
+  </critical>

package/src/tools/auto-generated-guard.ts

@@ -16,7 +16,7 @@ const CHECK_BYTE_COUNT = 1024;
 const HEADER_LINE_LIMIT = 40;
 
 const KNOWN_GENERATOR_PATTERN =
-	"(?:protoc(?:-gen-[\\w-]+)?|sqlc|buf|swagger(?:-codegen)?|openapi(?:-generator)?|grpc-gateway|mockery|stringer|easyjson|deepcopy-gen|defaulter-gen|conversion-gen|client-gen|lister-gen|informer-gen)";
+	"(?:protoc(?:-gen-[\\w-]+)?|sqlc|buf|swagger(?:-codegen)?|openapi(?:-generator)?|grpc-gateway|mockery|stringer|easyjson|deepcopy-gen|defaulter-gen|conversion-gen|client-gen|lister-gen|informer-gen|kysely-codegen)";
 
 /**
  * Strong marker patterns for generated-file headers.