fix: allow Claude Code to read skill seeds and design-system specs (#6) (#7)

* Allow Claude Code to read skill seeds and design-system specs (#6)

The skill body's preamble points the agent at absolute paths like
`<repo>/skills/guizang-ppt/assets/template.html`, but the agent's cwd
is `.od/projects/<id>/`. Without an explicit allowlist Claude Code
blocks Read on those paths and the user sees a permission error
mid-conversation.

Pass `SKILLS_DIR` and `DESIGN_SYSTEMS_DIR` through `buildArgs` and emit
them as `--add-dir` for Claude so the seed template, references, and
design-system DESIGN.md are all readable. Other agents ignore the
extra dirs (no equivalent flag).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* docs: add verification screenshot for issue #6 fix

Captures the agent successfully Read-ing skills/guizang-ppt/ side files
through the new --add-dir allowlist, confirming the permission error
from issue #6 is gone.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

daemon/agents.js

@@ -7,7 +7,12 @@ import path from 'node:path';
 const execFileP = promisify(execFile);
 
 // Each entry defines how to invoke the agent in non-interactive "one-shot" mode.
-// `buildArgs(prompt, imagePaths)` returns argv for the child process.
+// `buildArgs(prompt, imagePaths, extraAllowedDirs)` returns argv for the child
+// process. `extraAllowedDirs` is a list of absolute directories the agent must
+// be permitted to read files from (skill seeds, design-system specs) that live
+// outside the project cwd. Currently only Claude Code wires this through
+// (`--add-dir`); other agents either inherit broader access or run with cwd
+// boundaries we can't widen via flags.
 // `streamFormat` hints to the daemon how to interpret stdout:
 //   - 'claude-stream-json' : line-delimited JSON emitted by Claude Code's
 //     `--output-format stream-json`. Daemon parses it into typed events
@@ -19,14 +24,23 @@ export const AGENT_DEFS = [
     name: 'Claude Code',
     bin: 'claude',
     versionArgs: ['--version'],
-    buildArgs: (prompt) => [
+    buildArgs: (prompt, _imagePaths, extraAllowedDirs = []) => {
-      '-p',
+      const args = [
-      prompt,
+        '-p',
-      '--output-format',
+        prompt,
-      'stream-json',
+        '--output-format',
-      '--verbose',
+        'stream-json',
-      '--include-partial-messages',
+        '--verbose',
-    ],
+        '--include-partial-messages',
+      ];
+      const dirs = (extraAllowedDirs || []).filter(
+        (d) => typeof d === 'string' && d.length > 0,
+      );
+      if (dirs.length > 0) {
+        args.push('--add-dir', ...dirs);
+      }
+      return args;
+    },
     streamFormat: 'claude-stream-json',
   },
   {

daemon/server.js

@@ -769,7 +769,17 @@ export async function startServer({ port = 7456 } = {}) {
       safeImages.length ? `\n\n${safeImages.map((p) => `@${p}`).join(' ')}` : '',
     ].join('');
 
-    const args = def.buildArgs(composed, safeImages);
+    // Skill seeds (`skills/<id>/assets/template.html`) and design-system
+    // specs (`design-systems/<id>/DESIGN.md`) live outside the project cwd.
+    // The composed system prompt asks the agent to Read them via absolute
+    // paths in the skill-root preamble — without an explicit allowlist,
+    // Claude Code blocks those reads (issue #6: "no permission to read
+    // skills template"). We surface both roots so any agent that honours
+    // `--add-dir` can resolve those side files.
+    const extraAllowedDirs = [SKILLS_DIR, DESIGN_SYSTEMS_DIR].filter(
+      (d) => fs.existsSync(d),
+    );
+    const args = def.buildArgs(composed, safeImages, extraAllowedDirs);
 
     res.setHeader('Content-Type', 'text/event-stream');
     res.setHeader('Cache-Control', 'no-cache, no-transform');