docs: fix version references to v1 and prep v1.1.2 release

[csoceanu]

Summary

The README pointed users to a non-existent @v0 tag in the setup instructions. The actual release tags are v1.x (latest v1.1.1), so anyone copy-pasting the workflow snippet hits a "tag not found" error.

This PR aligns the docs and version metadata with the real v1 tag scheme, and preps a v1.1.2 release for the fixes already merged since v1.1.1.

Changes

• README.md — point users to @v1 (recommended moving tag) instead of @v0 in both the workflow snippet and the Versioning section; use @v1.1.2 as the exact-pin example.
• CHANGELOG.md — add the [1.1.2] section documenting the fixes landed since v1.1.1 (DOCS_SUBFOLDER normalization, stale-index cleanup, ls-tree CWD-relative pathspecs).
• pyproject.toml — bump version 0.1.0 → 1.1.2 (it had drifted at 0.1.0 across every v1.x release).

Notes

• No runtime behavior changes — all three files are docs/metadata; nothing reads the package version at runtime. Safe for @main users.
• After merge, pushing the v1.1.2 tag triggers the release workflow, which runs tests, creates the GitHub Release, and re-points the v1 major tag (currently stale) to this commit.

🤖 Generated with Claude Code

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 10:50 AM UTC · Completed 10:59 AM UTC

Commit: 804c2cf · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.15

[fullsend-ai-review]

Review

Findings

Low

• [documentation-consistency] CHANGELOG.md:10 — The changelog jumps from [0.1.0] to [1.1.2] without entries for intermediate releases v1.0.0–v1.1.1. The PR body explains that pyproject.toml had drifted at 0.1.0 across every v1.x release, and this PR improves the situation by establishing proper changelog practice going forward. Consider backfilling entries for prior v1.x releases in a follow-up.

[fullsend-ai-review]

[low] documentation-consistency

The changelog jumps from [0.1.0] to [1.1.2] without entries for intermediate releases v1.0.0-v1.1.1. The PR body explains that pyproject.toml had drifted at 0.1.0 across every v1.x release. This gap predates the PR and the PR improves the situation by establishing proper changelog practice starting with v1.1.2. Consider backfilling entries for prior v1.x releases in a follow-up.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 11:02 AM UTC · Completed 11:11 AM UTC

Commit: 804c2cf · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $3.24

[fullsend-ai-retro]

Retro: PR #83 — docs: fix version references to v1 and prep v1.1.2 release

Workflow timeline:

1. Human (csoceanu) + Claude Opus 4.8 locally authored a single-commit PR fixing @v0 → @v1 version references in README, adding a CHANGELOG v1.1.2 entry, and bumping pyproject.toml from 0.1.0 to 1.1.2.
2. Review agent (run 32960151929) dispatched at 10:49 UTC, completed at 10:59 UTC (~10 min). Model: claude-opus-4-6, effort: high, cost: $2.15.
3. Correctness sub-agent (opus, trivial scope): no issues. Style-conventions sub-agent (sonnet, trivial scope): 1 medium finding — CHANGELOG gap from [0.1.0] to [1.1.2] without intermediate entries.
4. Challenger sub-agent (opus) deliberately dismissed the finding, returning empty adjudicated_findings with reasoning that the gap predates the PR.
5. Orchestrator treated empty challenger result as a failure per process rules, fell back to pre-challenger findings, then used synthesis judgment to downgrade severity from medium → low.
6. Final verdict: APPROVED with 1 low advisory finding.
7. Fix agent dispatch (run 32961030967) — all stages skipped (approval, no fixes needed).
8. PR merged at 11:01 UTC, ~13 minutes after opening.

Overall assessment: The pipeline performed well — correct approval, reasonable finding, no rework, quick merge. Two improvement opportunities identified in the challenger-orchestrator interaction pattern.

Agents repo: fullsend-ai/agents@v0.37.0 (commit dc7c805632eb).

Proposals skipped (target repo not allowed)

File manually or update create_issues.allow_targets in config.yaml:

• Challenger should distinguish deliberate full-dismissal from malfunction (fullsend-ai/agents)
• Consider skipping challenger stage for trivial PRs with only low-severity findings (fullsend-ai/agents)