Add agent catalog modal

[klopez4212]

Summary

• Replace the catalog picker with a two-pane Agent Catalog modal that keeps the list on the left and full agent details on the right.
• Add local Share agent controls with catalog visibility and export moved out of the row action menu.
• Seed more inactive built-in catalog agents and keep custom agents visible in My Agents.

Screenshots

Agent Catalog

Share Agent

Checks

• git diff --check origin/main..HEAD
• cd desktop && ./node_modules/.bin/tsc --noEmit
• cd desktop && node scripts/check-px-text.mjs
• cd desktop && node --import ./test-loader.mjs --experimental-strip-types --test src/features/agents/lib/catalog.test.mjs
• . ./bin/activate-hermit && cargo test --manifest-path desktop/src-tauri/Cargo.toml managed_agents::personas::tests --lib
• cd desktop && ./node_modules/.bin/vite build
• cd desktop && ./node_modules/.bin/playwright test --project=integration agents.spec.ts

Note: pre-commit/pre-push hooks and pnpm ... biome check were attempted but blocked on pnpm install registry retries, so commit/push used --no-verify after the checks above passed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d9e848bc3e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

buzz/crates/buzz-agent/src/builtin.rs

Lines 203 to 205 in d885698

	content: vec![ToolResultContent::Text(format!(
	"# Loaded: {}/{}\n\n{}\n\n---\nFile loaded into context.",
	skill_name, rel_path_owned, content

Cap supporting-file load_skill output

The main SKILL.md load path caps returned text at MAX_SKILL_BODY_BYTES, but this supporting-file path interpolates the entire file into a built-in tool result. Because load_skill is handled inline and append_results stores the result without the MCP truncation budget, a large references/*.md file can be appended wholesale to history and blow the next LLM request/context. Cap content before formatting it here.

---

buzz/crates/buzz-agent/src/hints.rs

Lines 174 to 176 in d885698

	// Use std::fs::metadata (follows symlinks) so symlinked subdirs and files
	// inside a skill directory are handled correctly.
	let ft = match std::fs::metadata(&path) {

Avoid recursing through symlink cycles

Following symlinks here without tracking visited canonical directories lets a repo-provided skill hang startup with a symlink cycle such as .agents/skills/foo/references/loop -> ../references; path.join("SKILL.md") is false for that directory, so the recursive walk keeps descending until it overflows or stalls while building the hint list. Either avoid following symlinked directories or keep a visited set/depth limit.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 693ff02e89

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 564e1b2b40

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".