{
"tool": "edit",
"dual_mode": true,
"selection": "session_effective_hashline",
"arms": {
"legacy": {
"arm": "legacy",
"description": "Edit a file by finding and replacing text, or by targeting named symbols. To write or overwrite a whole file, use the `write` tool — `edit` requires an explicit edit mode and will not silently overwrite a file from `content` alone.",
"schema": {
"$schema": "https://json-schema.org/draft/2020-12/schema",
"type": "object",
"properties": {
"filePath": {
"description": "Path to the file to edit (absolute or relative to project root)",
"type": "string"
},
"symbol": {
"description": "Named symbol to replace (function, class, type)",
"type": "string"
},
"content": {
"description": "Replacement content for symbol mode. For whole-file writes, use the `write` tool.",
"type": "string"
},
"appendContent": {
"description": "Text to append to the end of path; creates the file if needed",
"type": "string"
},
"edits": {
"description": "Batch edits — non-empty array of { oldString, newString }, { oldString, newString, replaceAll: true }, or { startLine, endLine, content } objects",
"minItems": 1,
"type": "array",
"items": {
"type": "object",
"properties": {
"oldString": {
"description": "Text to find for a batch find/replace edit",
"type": "string"
},
"newString": {
"description": "Replacement text for a batch find/replace edit",
"type": "string"
},
"replaceAll": {
"description": "Replace every occurrence for this batch item",
"type": "boolean"
},
"occurrence": {
"description": "1-based occurrence for this batch item (1 = first match)",
"type": "integer",
"minimum": 1
},
"startLine": {
"description": "1-based start line for a batch line-range edit",
"type": "integer",
"minimum": 1
},
"endLine": {
"description": "1-based end line for a batch line-range edit",
"type": "integer",
"minimum": 1
},
"content": {
"description": "Replacement text for a batch line-range edit",
"type": "string"
}
}
}
}
},
"required": [
"filePath"
],
"description": "Edit a file by finding and replacing text, or by targeting named symbols. To write or overwrite a whole file, use the `write` tool — `edit` requires an explicit edit mode and will not silently overwrite a file from `content` alone."
},
"supports_tool": true,
"hoisted": true,
"lane": "mutation",
"command": "edit"
},
"hashline": {
"arm": "hashline",
"description": "Apply a hashline patch. Arguments are exactly `{patch}` where `patch` is a non-empty string. Server-owned preview control is outside this schema.\n\nQuick reference:\n- Header: `[path#TAG]`; TAG is exactly four hexadecimal digits from a current tagged read. Read every addressed row and gap boundary; REM and MV require a whole-file tagged read. Re-read after an edit before chaining: an edit-response tag can retain only changed context.\n- Same canonical path: multiple sections compose in patch order against pre-request coordinates.\n- Addresses: `0` (BOF), `N` (one line), `N.=M` (range; `N..=M`/`N..M` also work), `<N`/`>N` (gap before/after), `N*`/`<N*`/`>N*` (block), and `$`/`$-K` (EOF-relative). A plain `N` PUT replaces; use `<N` or `>N` to insert.\n- PUT text: `PUT <address>:` followed by one or more `+` body rows (`+` alone is blank). A final patch newline is allowed. PUT without `:` copies `@name` (or the anonymous register) and takes no body; names use `@` plus ASCII letters, digits, `_`, or `-`.\n- CUT: `CUT <address> [@name]`. REM: bare `REM` only, removing the whole file. MV: `MV <destination>` (one whitespace-free path, optional matching quotes), once and after any line operations. `*** Begin Patch`/`*** End Patch` is an optional envelope.",
"schema": {
"$schema": "https://json-schema.org/draft/2020-12/schema",
"type": "object",
"additionalProperties": false,
"properties": {
"patch": {
"type": "string",
"minLength": 1,
"description": "Hashline patch text with one or more [path#TAG] sections and PUT/CUT/REM/MV operations"
}
},
"required": [
"patch"
],
"description": "Apply a hashline patch. Arguments are exactly `{patch}` where `patch` is a non-empty string. Server-owned preview control is outside this schema.\n\nQuick reference:\n- Header: `[path#TAG]`; TAG is exactly four hexadecimal digits from a current tagged read. Read every addressed row and gap boundary; REM and MV require a whole-file tagged read. Re-read after an edit before chaining: an edit-response tag can retain only changed context.\n- Same canonical path: multiple sections compose in patch order against pre-request coordinates.\n- Addresses: `0` (BOF), `N` (one line), `N.=M` (range; `N..=M`/`N..M` also work), `<N`/`>N` (gap before/after), `N*`/`<N*`/`>N*` (block), and `$`/`$-K` (EOF-relative). A plain `N` PUT replaces; use `<N` or `>N` to insert.\n- PUT text: `PUT <address>:` followed by one or more `+` body rows (`+` alone is blank). A final patch newline is allowed. PUT without `:` copies `@name` (or the anonymous register) and takes no body; names use `@` plus ASCII letters, digits, `_`, or `-`.\n- CUT: `CUT <address> [@name]`. REM: bare `REM` only, removing the whole file. MV: `MV <destination>` (one whitespace-free path, optional matching quotes), once and after any line operations. `*** Begin Patch`/`*** End Patch` is an optional envelope."
},
"supports_tool": true,
"hoisted": true,
"lane": "mutation",
"command": "hashline_edit",
"preflight_command": "hashline_preflight"
}
},
"invariant": "a session never exposes both edit schemas"
}