Labali Weibo Download User Posts

NEVER

• Never delete or discard partially downloaded files on URL failure — keep all successfully downloaded content.
• Never report success unless at least one post was extracted and the output folder was created with posts.json and user.md.
• Never report success based on action completion alone — verify the output folder and metadata files exist.

Success Criteria

A run is successful only when all conditions hold:

1. A user output folder is created under target local directory.
2. Folder naming format is <timestamp>-<user_slug>.
3. posts.json and user.md are generated.
4. At least one post is extracted from timeline.
5. Per-post folders contain post.md and downloaded media when available.
6. URL output and logs retain canonical user URL form.

Runtime Inputs

Use skill.yaml as the source of truth for input schema.

Operational Mode

• Default mode: guided browser flow + semantic extraction + authenticated media download.
• Startup guidance:
  • launch/reuse Chrome by the unified open -na "Google Chrome" CDP command,
  • connect via CDP port,
  • open target user page,
  • check login status,
  • if required, guide user to complete manual login in the same window.
• Input guidance:
  • if user_url is missing, prompt interactively,
  • if output_dir is missing, prompt interactively with default.
• Pagination guidance:
  • scroll timeline,
  • click visible expansion controls,
  • stop when no more content appears or explicit end markers are present.

If failure handling or extraction decisions are unclear, load references/architecture.md. If extraction strategy for a specific content type is unclear, load references/plan.md.

Resources

• Architecture and standards: references/architecture.md
• Workflow map and extraction plan: references/plan.md
• Shared runtime and downloader helpers: scripts/core.ts
• Main orchestration: scripts/executor.ts
• CLI entry: scripts/run.ts
• Regression checks: tests/test_regression.sh