feat(session): persist model tier overrides per session

Store per-session config in SQLite and route /model and /reset through command fast-paths so channel sessions keep independent model selection across reconnects and restarts.

src/config/schema.test.ts

@@ -150,6 +150,35 @@ describe('configSchema — skills watcher', () => {
     const result = configSchema.parse(minimalConfig);
     expect(result.skills.load.watch).toBe(false);
     expect(result.skills.load.watch_debounce_ms).toBe(250);
+    expect(result.skills.installation_execution).toBe('disabled');
+    expect(result.skills.allow_shell_runner).toBe(false);
+    expect(result.skills.shell_runner_allowlist).toEqual([]);
+    expect(result.skills.shell_runner_governance.owner).toBeUndefined();
+    expect(result.skills.shell_runner_governance.review_cadence_days).toBe(7);
+    expect(result.skills.shell_runner_governance.promotion_min_success_rate).toBe(0.9);
+  });
+
+  it('accepts explicit installation execution policy', () => {
+    const enabled = configSchema.parse({
+      ...minimalConfig,
+      skills: {
+        installation_execution: 'enabled',
+        allow_shell_runner: true,
+        shell_runner_allowlist: ['npm install*'],
+        shell_runner_governance: {
+          owner: 'skills-team',
+          review_cadence_days: 14,
+          promotion_min_success_rate: 0.95,
+        },
+      },
+    });
+
+    expect(enabled.skills.installation_execution).toBe('enabled');
+    expect(enabled.skills.allow_shell_runner).toBe(true);
+    expect(enabled.skills.shell_runner_allowlist).toEqual(['npm install*']);
+    expect(enabled.skills.shell_runner_governance.owner).toBe('skills-team');
+    expect(enabled.skills.shell_runner_governance.review_cadence_days).toBe(14);
+    expect(enabled.skills.shell_runner_governance.promotion_min_success_rate).toBe(0.95);
   });
 
   it('accepts explicit watcher settings', () => {

src/config/schema.ts

@@ -108,6 +108,15 @@ const skillsLoadSchema = z.object({
   watch_debounce_ms: z.number().min(10).max(10_000).default(250),
 }).default({});
 
+const skillsShellRunnerGovernanceSchema = z.object({
+  /** Responsible owner for shell-runner allowlist decisions. */
+  owner: z.string().min(1).optional(),
+  /** Review cadence for allowlist + rollout status checks. */
+  review_cadence_days: z.number().min(1).max(90).default(7),
+  /** Minimum success rate required before broader rollout. */
+  promotion_min_success_rate: z.number().min(0).max(1).default(0.9),
+}).default({});
+
 const skillsSchema = z.object({
   /** Directory for user-created workspace skills. */
   workspace_dir: z.string().optional(),
@@ -115,6 +124,14 @@ const skillsSchema = z.object({
   managed_dir: z.string().optional(),
   /** Directory for bundled skills shipped with Flynn. */
   bundled_dir: z.string().optional(),
+  /** Global policy gate for installer command execution. */
+  installation_execution: z.enum(['disabled', 'enabled']).default('disabled'),
+  /** Allow use of the shell runner for installer commands. */
+  allow_shell_runner: z.boolean().default(false),
+  /** Allowlist patterns for shell runner commands (supports '*' wildcard). */
+  shell_runner_allowlist: z.array(z.string()).default([]),
+  /** Governance controls for shell-runner rollout decisions. */
+  shell_runner_governance: skillsShellRunnerGovernanceSchema,
   /** Skills watcher settings. */
   load: skillsLoadSchema,
 }).default({});