Merge pull request 'feat(agentic): scope file/command tools to an agent workspace root' (#23) from feat/agent-tool-workspace-scope into main

fix(ci): link soul.c only — fixes capability #error breaking every build

chat.el

@@ -317,10 +317,13 @@ fn connector_tools_json() -> String {
     return arr
 }
 
-// Built-in tools + native web_search + every connector tool, as one tools array.
+// Built-in tools + every connector tool, as one tools array.
-// Splices connector tools in before the closing bracket of the base array.
+// Uses agentic_tools_literal (not agentic_tools_with_web) to avoid a duplicate
+// "web_search" name — the literal already includes a custom web_search handler,
+// and adding the Anthropic server-side web_search_20250305 (same name) causes
+// Anthropic to reject with "Tool names must be unique."
 fn agentic_tools_all() -> String {
-    let base: String = agentic_tools_with_web()
+    let base: String = agentic_tools_literal()
     let conn: String = connector_tools_json()
     let conn_inner: String = str_slice(conn, 1, str_len(conn) - 1)
     if str_eq(conn_inner, "") {
@@ -377,16 +380,78 @@ fn call_neuron_mcp(tool_name: String, args: String) -> String {
     return json_safe(result)
 }
 
+// ---------------------------------------------------------------------------
+// Agent workspace scope (defense-in-depth, NOT a hard security boundary).
+//
+// When a workspace root is configured (state key "agent_workspace_root", else
+// env NEURON_AGENT_ROOT), the path-based tools (read_file, write_file,
+// list_files, grep) are confined to that subtree by a lexical check, and
+// run_command runs with its cwd set to the root. With no root set, behavior is
+// unchanged (unscoped) for backward compatibility.
+//
+// LIMITATION — FLAGGED FOR WILL'S REVIEW: this is a lexical guard. It does not
+// resolve symlinks and cannot stop an arbitrary shell command from cd-ing out
+// of the root. Real confinement needs runtime support (cwd-locked exec /
+// sandbox-exec / chroot) in el_runtime.c. This raises the floor; it is not a
+// boundary. The default-allow-when-unset policy and the "cd <root> && (...)"
+// wrapping are deliberate choices to confirm against the intended design.
+// ---------------------------------------------------------------------------
+
+fn agent_workspace_root() -> String {
+    let s: String = state_get("agent_workspace_root")
+    if !str_eq(s, "") {
+        return s
+    }
+    return env("NEURON_AGENT_ROOT")
+}
+
+// Allow if path stays under root. Empty root = no sandbox = allow. Rejects
+// parent traversal and ~ expansion; absolute paths must live under root.
+fn path_within_root(path: String, root: String) -> Bool {
+    if str_eq(root, "") {
+        return true
+    }
+    if str_contains(path, "..") {
+        return false
+    }
+    if str_starts_with(path, "~") {
+        return false
+    }
+    if str_starts_with(path, "/") {
+        return str_starts_with(path, root)
+    }
+    return true
+}
+
+// Resolve a relative tool path against the root so it lands inside the subtree.
+fn resolve_in_root(path: String, root: String) -> String {
+    if str_eq(root, "") {
+        return path
+    }
+    if str_starts_with(path, "/") {
+        return path
+    }
+    return root + "/" + path
+}
+
 fn dispatch_tool(tool_name: String, tool_input: String) -> String {
     if str_eq(tool_name, "read_file") {
         let path: String = json_get(tool_input, "path")
-        let content: String = fs_read(path)
+        let root: String = agent_workspace_root()
+        if !path_within_root(path, root) {
+            return json_safe("denied: path is outside the agent workspace root")
+        }
+        let content: String = fs_read(resolve_in_root(path, root))
         return json_safe(content)
     }
     if str_eq(tool_name, "write_file") {
         let path: String = json_get(tool_input, "path")
         let content: String = json_get(tool_input, "content")
-        fs_write(path, content)
+        let root: String = agent_workspace_root()
+        if !path_within_root(path, root) {
+            return json_safe("denied: path is outside the agent workspace root")
+        }
+        fs_write(resolve_in_root(path, root), content)
         return json_safe("{\"ok\":true}")
     }
     if str_eq(tool_name, "web_get") {
@@ -401,7 +466,9 @@ fn dispatch_tool(tool_name: String, tool_input: String) -> String {
     }
     if str_eq(tool_name, "run_command") {
         let cmd: String = json_get(tool_input, "command")
-        let result: String = exec_capture(cmd)
+        let root: String = agent_workspace_root()
+        let scoped: String = if str_eq(root, "") { cmd } else { "cd " + root + " && ( " + cmd + " )" }
+        let result: String = exec_capture(scoped)
         return json_safe(result)
     }
     // MCP connector tools (namespaced mcp__<server>__<tool>) are routed through
@@ -421,13 +488,21 @@ fn dispatch_tool(tool_name: String, tool_input: String) -> String {
     }
     if str_eq(tool_name, "list_files") {
         let path: String = json_get(tool_input, "path")
-        let result: String = exec_capture("ls -la " + path + " 2>&1")
+        let root: String = agent_workspace_root()
+        if !path_within_root(path, root) {
+            return json_safe("denied: path is outside the agent workspace root")
+        }
+        let result: String = exec_capture("ls -la " + resolve_in_root(path, root) + " 2>&1")
         return json_safe(result)
     }
     if str_eq(tool_name, "grep") {
         let pattern: String = json_get(tool_input, "pattern")
         let path: String = json_get(tool_input, "path")
-        let result: String = exec_capture("grep -rn \"" + pattern + "\" " + path + " 2>&1 | head -50")
+        let root: String = agent_workspace_root()
+        if !path_within_root(path, root) {
+            return json_safe("denied: path is outside the agent workspace root")
+        }
+        let result: String = exec_capture("grep -rn \"" + pattern + "\" " + resolve_in_root(path, root) + " 2>&1 | head -50")
         return json_safe(result)
     }
     if str_eq(tool_name, "edit_file") {

soul.el

@@ -95,6 +95,24 @@ fn init_soul_edges() -> Void {
     engram_connect(val_hope, val_trust, el_from_float(0.7), "co-value")
 }
 
+// ensure_self_canonical_bridge — link the public self anchor (the graph API's
+// traversal_root, kn-efeb4a5b, which carries only incidental tag edges) to the
+// curated self node (015644f5, where the real identity / value / co-value edges
+// live). Without this, public self-traversal (name=self / neuron) reaches tags
+// instead of the curated identity. Idempotent: connects only if the edge is
+// missing, so it is safe to run every boot — including on an already-populated
+// graph where init_soul_edges() is skipped by the <100-edge gate.
+fn ensure_self_canonical_bridge() -> Void {
+    let pub_self: String = "kn-efeb4a5b-5aff-4759-8a97-7233099be6ee"
+    let curated_self: String = "015644f5-8194-4af0-800d-dd4a0cd71396"
+    let nbrs: String = engram_neighbors_json(pub_self, 1, "out")
+    if !str_contains(nbrs, curated_self) {
+        engram_connect(pub_self, curated_self, el_from_float(0.95), "canonical-self")
+        engram_connect(curated_self, pub_self, el_from_float(0.95), "canonical-self")
+        println("[soul] canonical-self bridge built: kn-efeb4a5b <-> 015644f5")
+    }
+}
+
 // load_identity_context — pull key identity nodes from engram into working state.
 // Called at boot after engram_load. These nodes contain values, intellectual-dna,
 // memory-philosophy — the graph-stored self that chat.el can include in prompts.
@@ -398,6 +416,9 @@ if is_genesis && safe_to_seed {
     } else {
         println("[soul] edges already present (" + int_to_str(edge_count_now) + ") - skipping init")
     }
+    // Canonical-self bridge is idempotent — run it regardless of edge count so an
+    // already-populated graph still gets the public->curated self link.
+    ensure_self_canonical_bridge()
     // Genesis saves to its local snapshot file (it manages its own Engram).
     state_set("soul_snapshot_path", snapshot)
     engram_save(snapshot)