Stop Codex from satisfying a request by copying an unrelated pre-existing
image from generated_images instead of generating a new one. Verification
now requires a PNG in this thread's generated_images dir (or an image_gen
stream item, kept as a forward-compatible signal), and the spawned-agent
instruction forbids reading or reusing history images before image_gen is
called. Removes the findCpToTarget fallback that allowed the shortcut.
Validated against a real codex exec run: image_gen leaves no stream item,
so the filesystem check is the load-bearing signal; locked in via a
condensed real-stream regression fixture.
Closes#185
parseYouTubeVideoId handled watch, youtu.be, /shorts/ and /live/ URLs but
not the common /embed/<id> player form, so embed links returned null and the
YouTube adapter treated them as "no document". Add an /embed/ branch
mirroring the existing /shorts/ and /live/ handling, with a regression test.
Obsidian creates markdown image links with URL-encoded filenames
(e.g. `Pasted%20image%2020260524.png`) but the actual file has spaces.
Add `decodeURIComponent()` before path resolution in both baoyu-md's
shared `resolveImagePath()` and baoyu-post-to-x's independent version.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move the codex-imagegen wrapper out of scripts/ into its own workspace
package alongside baoyu-md, baoyu-chrome-cdp, and baoyu-fetch. Drop the
bash shim — src/main.ts now carries a `#\!/usr/bin/env bun` shebang and
serves as the sole entrypoint. Add scripts/sync-codex-imagegen.sh so
skills/baoyu-image-gen/scripts/codex-imagegen/ can be regenerated from
packages/baoyu-codex-imagegen/src/ for skill self-containment. Update
CLAUDE.md, docs/codex-imagegen-backend.md, and the CI workflow paths
accordingly.
Add cleanSummaryText() to baoyu-md package: decodes HTML entities (&, <, &#x..., etc.)
and strips HTML tags before using frontmatter description/summary as WeChat article digest
Add cookie sidecar to export/restore X session cookies across runs.
Implement graceful Browser.close via WebSocket before falling back to kill.
Auto-detect and clean stale Chrome profile lock artifacts on launch failure.
Wait for X session readiness (auth_token + ct0) before auto-continuing login flows.
- Remove quotes from CSS custom property regex character class so values containing quotes are fully stripped
- grace/simple themes now layer default CSS before their own rules
- Add tests for quoted property stripping and theme layering
findExistingChromeDebugPort callers (fetch_google_cookies_via_cdp) depend
on /json/version being available. TCP-only check could misclassify
non-CDP listeners as reusable, causing waitForChromeDebugPort to timeout
instead of falling back to launching a new Chrome. Restore isDebugPortReady
(HTTP) as the validator for this function; TCP-only check remains in
discoverRunningChromeDebugPort for Chrome 146 approval mode.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Chrome 146+ blocks all HTTP endpoints (/json/version) in approval mode.
Read DevToolsActivePort ws path directly and use TCP port check instead
of HTTP discovery. Add WebSocket connect retry loop for approval dialog.
Unify findExistingChromeDebugPort to use the same mechanism.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add ability to discover and connect to an already-running Chrome browser
(v144+) for cookie extraction, avoiding the need to launch a new window
and re-login. Uses Chrome's DevToolsActivePort from default user data
directories and process scanning as discovery mechanisms.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move test files from tests/ directory to colocate with source code,
convert from .mjs to .ts using tsx runner, add workspaces and npm
cache to CI workflow.
Jim Liu 宝玉
Syed Osama Ali Shah
Chao Zheng
Alec Jiang
shixy