docs: add build-from-source instructions to README install section

[gxmiranda]

Summary

Adds a "Build from Source" subsection to the README ## Install
section, filling a documentation gap tracked by #90. The replicator
README listed Homebrew, go install, and binary download but omitted
build-from-source instructions — inconsistent with gaze and dewey
which both document this path.

The new subsection documents prerequisites (Go 1.25+, make), clone +
build commands (make build), output location (bin/replicator), and
a verification step (bin/replicator version). Uses make build
rather than bare go build because the Makefile injects version,
commit, and date via ldflags.

Closes #90

How to Test

1. Clone the repo and run make build
2. Run bin/replicator version — should show version, commit, and
   build date (not dev/unknown)
3. Compare the new README subsection against gaze and dewey
   build-from-source sections for structural consistency

How to Demo

Read the README ## Install section — the new "Build from Source"
subsection should appear after "Binary Download" with a code block
containing clone, cd, and make build commands, followed by a
prerequisites note.

Key Files Changed

• README.md — Added "Build from Source" subsection (10 lines)
• openspec/changes/build-from-source-readme/ — OpenSpec change
  artifacts (proposal, design, specs, tasks)

This PR was generated by /uf.finale (AI-assisted).

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 12:25 PM UTC · Completed 12:37 PM UTC

Commit: e33020e · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $3.39

[fullsend-ai-review]

Risk Assessment: low (1/5)

Details

Documentation-only PR adding build-from-source instructions to README and OpenSpec planning artifacts. No source code, tests, CI, dependencies, or security-sensitive files are touched. Linked issue scope matches the PR exactly with all acceptance criteria covered.

[fullsend-ai-review]

Looks good to me

[em-redhat]

PR Review: #104 — docs: add build-from-source instructions to README install section

Summary: Clean, well-scoped documentation-only PR. All 5 spec requirements satisfied, all 3 issue #90 acceptance criteria met, CI green, no code touched.

Findings (all non-blocking):

• [LOW] Alignment: Prerequisites documented as inline prose rather than a bulleted list before the code block — differs slightly from gaze/dewey pattern but information is complete and accurate.
• [LOW] Alignment: tasks.md references external repo line numbers (gaze/dewey) that can drift independently. No user-facing impact.

Constitution: Compliant. Full OpenSpec workflow, correct branch naming (opsx/), Conventional Commits format.

CI: All checks PASS or SKIPPED. No failures.

Verdict: APPROVE — ready to merge.

This review was generated by /uf.review-pr (AI-assisted).

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 11:53 AM UTC · Completed 12:03 PM UTC

Commit: 9c087ac · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.99

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 11:55 AM UTC · Completed 12:05 PM UTC

Commit: 9c087ac · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.94

[fullsend-ai-review]

Review skipped — this PR is already merged.

The /fs-review command only reviews open PRs/MRs.

_{Posted by fullsend post-review check}

[fullsend-ai-retro]

Retro: PR #104 — docs: add build-from-source instructions

Workflow quality: Good. This was a clean, docs-only PR adding build-from-source instructions to the README and closing #90. Zero rework, single commit, correct approvals from both the review agent and human reviewer. No meaningful improvement proposals are warranted.

Timeline

1. 2026-08-24 — Issue #90 created by gxmiranda
2. 2026-09-14 12:23 — PR #104 opened via /uf.finale (6 files, 230 additions — 10 lines of README + OpenSpec artifacts)
3. 2026-09-14 12:25–12:37 — Fullsend review agent ran (run 34843198115): opus/high, $3.39, approved with no findings
4. 2026-09-17 11:48 — Human reviewer em-redhat approved via /uf.review-pr with two LOW non-blocking findings
5. 2026-09-17 11:52 — Second review triggered on rebased commit (run 35217974786)
6. 2026-09-17 11:54 — PR merged before second review completed

Observations

• Review quality: The agent correctly approved a clean docs-only PR with no false positives. The human reviewer's two LOW findings (formatting consistency with sibling repos, brittle line-number references in tasks.md) were reasonable polish but non-blocking — the agent's approval was appropriate.
• Cost: $3.39 (opus, high effort) for reviewing 10 lines of README changes + OpenSpec planning artifacts. The risk assessment correctly classified the PR as low (1/5).
• Wasted second review: A re-review was triggered at 11:52 after a rebase, but the PR was merged at 11:54 before it completed — those tokens were wasted. This is a human timing issue (merging before the agent finishes), not a routing bug.

Evidence for existing issues (no new proposals needed)

• agents#373 and fullsend#6891: This PR is a concrete data point — $3.39 opus/high for a trivially-scoped docs-only change. A pre-flight scope heuristic or skipping code-analysis dimensions would have reduced cost significantly while producing the same correct approval.
• agents#427: The review agent competently handled this docs-only PR — correct approval, accurate risk assessment (low 1/5), no false positives. Supports the autonomy evidence that the agent handles docs-only PRs well.
• agents#649: This retro itself ran on a trivially clean workflow with no actionable findings, supporting the case for early-exit on mechanical/trivially-handled PRs.