Skip to content

docs: milestone sync — the "what Orphus authors" maps missed an arc of first-party code - #104

Open
kelvincushman wants to merge 2 commits into
mainfrom
docs/milestone-sync-harness-arc
Open

docs: milestone sync — the "what Orphus authors" maps missed an arc of first-party code#104
kelvincushman wants to merge 2 commits into
mainfrom
docs/milestone-sync-harness-arc

Conversation

@kelvincushman

Copy link
Copy Markdown
Owner

The milestone docs sync AGENTS.md asks for after a feature arc lands, run against main at ab5070d (#101 merged).

What was actually wrong

Two files tell a reader which half of this tree is vendored and which half Orphus wrote. Both still said roundtable and fleet. After #101 that is wrong in a way that changes behaviour: a contributor reading docs/architecture.md would conclude the browser extension and the capability boundary are Atomic's, and either report a bug upstream or leave the files alone.

File Was Now
AGENTS.md Package map listed eight @orphus/* packages; @orphus/transcribe was not one of them Added, under the first-party heading rather than "inherited from Atomic"
README.md Tree omitted packages/transcribe/; coding-agent labelled simply "(Atomic-derived)" Both corrected
docs/architecture.md "What Orphus authors" table: roundtable, fleet, their tests Plus transcribe and the four coding-agent subsystems, by real path
docs/README.md Same claim in prose Matches, and links the path table as the thing to check

Two findings beyond the obvious

@orphus/transcribe was a workspace package no agent-facing map listed. Its entry now says the quiet part explicitly — not bundled, fails closed, and that is the intended state, not a guard to remove. Without it, the next agent to notice the package is unregistered has every reason to "fix" it.

"A question about the harness underneath is usually answered upstream." That sentence in docs/README.md predates harness.md, which documents Orphus's own capability boundary and session records. The word now points at both halves at once, and the sentence sent readers to exactly the wrong one. Rewritten, with the ambiguity called out rather than left for someone to trip over.

Verification

npm run check      # biome 2454 files, tsc --noEmit, shrinkwrap — green

Every path named in the new table was checked to exist before it was written — the first draft said packages/coding-agent/src/harness/, which does not exist; the real locations are core/{capabilities,replay}/, core/provider-audit.ts, cli/inspect-runtime.ts, extensions/browser/, and core/terminal/termdom-*. The test/unit/{harness,browser,terminal,transcribe}-* glob matches 16 files.

No changelog entry: documentation-only, no shipped behaviour changed.


Generated by Claude Code

The four-stage harness arc landed as one PR, and the maps that tell a reader
which half of this tree is vendored did not move with it. Both said Orphus
authors roundtable and fleet — so a contributor reading them would conclude the
browser extension and the capability boundary are upstream's, and either report
a bug to Atomic or leave the files alone.

- AGENTS.md gained @orphus/transcribe, which was a workspace package no agent
  map listed. Its entry says the quiet part: not bundled, fails closed, and that
  is the intended state rather than a guard to remove.
- README's tree gained packages/transcribe and stops calling coding-agent
  wholly Atomic-derived.
- architecture.md's path table gained transcribe and the four coding-agent
  subsystems, with the real paths (core/{capabilities,replay}, provider-audit.ts,
  cli/inspect-runtime.ts, extensions/browser, core/terminal/termdom-*).
- docs/README.md's two-halves note matches, and no longer sends "a question
  about the harness underneath" upstream — harness.md is about Orphus's own
  capability boundary, which is the opposite of what that sentence implied.

No changelog entry: documentation-only, no shipped behaviour changes.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WSJaWPmbk52kSjoJi9TQ1a
@coderabbitai

coderabbitai Bot commented Aug 19, 2026

Copy link
Copy Markdown

Warning

Review limit reached

You’ve reached a temporary PR review limit under our Fair Usage Limits Policy.

Your current included review allowance is based on your included PR review attempts over the past 7 days.

Next review available in: 23 minutes

Limit details: You’ve used the included review currently available. Your 64 included PR review attempts over the past 7 days set your current allowance at 1 review per hour.

Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab.

How can I continue?

Wait for the limit to reset, then comment @coderabbitai review or push new commits to the PR.

An organization admin can change what happens after included review limits in Billing.

How do review limits work?

CodeRabbit enforces per-developer PR review limits within each organization.

For paid Pro and Pro+ reviews, CodeRabbit uses a developer's included PR review attempts over the past 7 days to set the current hourly allowance. At typical activity levels, the full plan allowance applies. Higher sustained activity can lower the allowance until earlier attempts leave the 7-day window.

Please refer docs for additional details.

Review details
⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 129e2e59-2871-4329-8f4e-833d8cd22d9d

📥 Commits

Reviewing files that changed from the base of the PR and between 6cc4763 and 22f39aa.

📒 Files selected for processing (4)
  • AGENTS.md
  • README.md
  • docs/README.md
  • docs/architecture.md

Comment @coderabbitai help to get the list of available commands.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants