Seed collaboration showcase fixtures - #183
Conversation
The development seed stopped short of the newest features: nothing demonstrated agent attribution, comment threads, citation back matter, or the library audit trail, so demos and screenshots needed one-off scripts. Adds a collab-showcase document plus fixtures for each, following the existing deterministic, never-clobber-local-edits pattern. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 726d061531
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".
| plan_version: plan.current_plan_version, | ||
| created_by_user: user, | ||
| anchor_text: anchor, | ||
| status: "pending" |
There was a problem hiding this comment.
When this helper creates the agent remark above, user is Priya—the showcase plan's author—but the unconditional pending status renders it as reviewer feedback and asks Priya to accept her own comment. Derive the initial status from plan ownership or explicitly pass todo so the seed demonstrates the documented thread lifecycle.
AGENTS.md reference: AGENTS.md:L132-L134
Useful? React with 👍 / 👎.
1 participant
Why
The development seed stopped short of the features that landed over the last few weeks: nothing demonstrated agent attribution (#178/#181), comment threads in any state, citation back matter (#177), folder descriptions or the library audit trail (#176). Demos, screenshots, and manual testing all required hand-crafted one-off scripts — the "Claude (via Hampton)" screenshot for this week's #coplan update was staged exactly that way.
What
collab-showcasedocument ("Search latency: cutting p95 with a two-tier cache") exercising the reference system end-to-end: footnote citations with source links, explicit[§3](#section-3)numbered-section links, a tagged code fence, and prose the comment fixtures anchor toPlans::ReplaceContent— history renders "Claude (via …)" with fullapi_tokenprovenance, backed by a seeded root token carrying harness metadatatarget_plan, and one explicit curated repository referencerun_id, each audit event carrying agent name and token provenance)Risk Assessment
Low — development/test seed data only; no production code paths change. Fixtures follow the existing deterministic, never-clobber-local-edits pattern: anchors that no longer match after local edits skip with a warning instead of failing the seed.
References
Generated with Claude Code
Screenshots
The showcase document — citation superscripts, section links, comment highlights, TOC badges
Citation hover preview with source identity
References back matter — extracted citations, a resolved plan reference, and a curated repository
History with agent attribution