feat: align sweep with Phase 2 naming, add aggregate sweep summary

[pruiz]

Summary

Aligns the file-risk sweep tooling with Phase 2 naming conventions and adds an aggregate sweep summary step.

Changes

• Rename prompts/sweep.md → prompts/phase-2-sweep.md
• New prompts/phase-2-sweep-summary.md — aggregate sweep rollup prompt (consolidates per-file summaries, forbids new findings/hunting, writes runs/sweep-summary-*.md, prints to screen)
• Per-file sweeps write runs/phase-2-summary-sweep-<slug>-*.md (satisfies Phase 2 completion gate)
• find_latest_summary("2") now unconditionally excludes per-file sweep summaries from broad Phase 2 lookup
• find_latest_sweep_summary() added for runs/sweep-summary-*.md lookup
• codecome hints: Phase 2 uses filtered lookup; Sweep block surfaces aggregate rollup; removed old per-file sweep scan
• tools/run-sweep.py: invokes aggregate summary via opencode run after all per-file sweeps succeed (skipped on dry-run); only feeds fresh per-file summaries to the aggregate prompt
• README/docs updated for new prompt paths and artifact names; removed deprecated CODECOME_USE_WRAPPER references
• Tests added for sweep exclusion, find_latest_sweep_summary, aggregate summary prompt generation, stale-summary filtering

Follow-up commit (2cc850b) addresses review feedback

• Docstring now says "extend" not "override" (the implicit Phase 2 sweep exclusion cannot be overridden)
• exclude_patterns matching uses fnmatch for clarity
• Removed deprecated CODECOME_USE_WRAPPER branch from per-file sweeps; per-file runs always use the wrapper path
• Documented why aggregate rollup uses raw opencode run (not a phase-mode run, run-agent.py lacks non-phase utility support)
• Simplified redundant aggregate-summary guard
• Aggregate prompt now injects exact per-file summary paths and instructs the model to read only those (not all historical phase-2-summary-sweep-*.md)
• Fixed slug example in per-file sweep prompt to match slugify() output
• Timestamped aggregate prompt path instead of fixed name
• Cleaned CODECOME_USE_WRAPPER from README.md and docs/workflow.md

Artifact naming

File	Purpose
runs/phase-2-summary-*.md	Broad make phase-2
runs/phase-2-summary-sweep-<slug>-*.md	Per-file sweep (Phase 2 completion gate)
runs/sweep-summary-*.md	Aggregate sweep rollup (codecome hints Sweep block)
tmp/file-sweep-prompts/sweep-summary-YYYYMMDD-HHMMSS.md	Generated aggregate prompt

[coderabbitai]

Warning

Review limit reached

@pruiz, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 7 minutes and 40 seconds. Learn how PR review limits work.

Your organization has run out of usage credits. Purchase more credits in the billing tab to continue.

⌛ How to resolve this issue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans include higher PR review limits than trial, open-source, and free plans. In all cases, reviews become available again over time. During sustained high-volume PR review activity, CodeRabbit may temporarily slow when the next review becomes available.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 3fb5cea7-81de-4fa3-8806-3d52bcffaa43

📥 Commits

Reviewing files that changed from the base of the PR and between cb84c51 and 2cc850b.

📒 Files selected for processing (12)

• .project/phase-2-sweep-alignment-plan.md
• README.md
• docs/file-risk-sweeps.md
• docs/workflow.md
• prompts/phase-2-sweep-summary.md
• prompts/phase-2-sweep.md
• tests/test_phases_completion.py
• tests/test_run_summary_questions.py
• tests/test_run_sweep.py
• tools/codecome.py
• tools/codecome/run_summary_questions.py
• tools/run-sweep.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/phase-2-sweep-alignment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

Coverage Report

Metric	Value
Line Coverage	75.8%
Lines Covered	0 / 0

Download detailed HTML coverage reports per OS/Python from the workflow artifacts.

Generated by pytest-cov on 2026-06-12T19:08:41.737Z

[Copilot]

Pull request overview

This PR aligns the file-risk sweep workflow with Phase 2 naming conventions and adds an aggregate “sweep rollup” step so operators can review one consolidated summary after per-file sweeps complete.

Changes:

• Renames/updates sweep prompt to Phase 2 sweep mode and switches per-file sweep summaries to runs/phase-2-summary-sweep-<slug>-*.md (so Phase 2 completion gates are satisfied).
• Adds an aggregate sweep summary prompt (prompts/phase-2-sweep-summary.md) and runner orchestration to produce runs/sweep-summary-*.md after successful per-file sweeps.
• Updates codecome hints and summary lookup helpers/tests so broad Phase 2 summaries exclude per-file sweep summaries, while Sweep uses the aggregate rollup.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tools/run-sweep.py	Uses Phase 2 sweep prompt and runs an aggregate sweep rollup step after per-file sweeps.
tools/codecome/run_summary_questions.py	Adds Phase 2 sweep-summary exclusion to broad summary lookup + adds find_latest_sweep_summary().
tools/codecome.py	Updates codecome hints to show Sweep from aggregate sweep-summary-*.md only.
tests/test_run_sweep.py	Adds tests for new prompt paths and aggregate sweep summary prompt generation.
tests/test_run_summary_questions.py	Adds tests for Phase 2 sweep exclusion + find_latest_sweep_summary().
tests/test_phases_completion.py	Adds completion-gate tests verifying sweep uses Phase 2-style summary artifacts.
README.md	Updates sweep docs to reflect new prompt paths and artifact naming.
prompts/phase-2-sweep.md	Updates sweep prompt to Phase 2 naming and per-file sweep summary path.
prompts/phase-2-sweep-summary.md	New aggregate sweep rollup prompt (consolidation-only, no new findings).
docs/file-risk-sweeps.md	Updates sweep docs for Phase 2 sweep mode + aggregate rollup behavior and artifacts.
.project/phase-2-sweep-alignment-plan.md	Adds design/implementation plan documenting the sweep alignment approach.

Comments suppressed due to low confidence (1)

prompts/phase-2-sweep.md:162

• The example sweep summary filename uses upload-php (dot replaced with a hyphen), but the sweep runner’s slugify() preserves dots (e.g., upload.php). This mismatch can confuse operators/models about the expected <slug> format.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[greptile-apps]

Greptile Summary

This PR aligns the file-risk sweep tooling with Phase 2 naming conventions and adds an aggregate sweep summary step that consolidates per-file results into a single rollup after all sweeps complete.

• Rename prompts/sweep.md → prompts/phase-2-sweep.md; per-file sweep summaries now write to runs/phase-2-summary-sweep-<slug>-*.md, satisfying the Phase 2 completion gate.
• New prompts/phase-2-sweep-summary.md and run_sweep_summary() orchestrate a final opencode run rollup step that writes runs/sweep-summary-*.md; codecome hints surfaces this via a new Sweep block using find_latest_sweep_summary().
• find_latest_summary(\"2\") now unconditionally excludes per-file sweep summaries via fnmatch, preventing them from polluting the broad Phase 2 block in codecome hints.

Confidence Score: 5/5

Safe to merge; the naming alignment, lookup helpers, and docs are all internally consistent, and the test coverage is solid.

The core logic is correct and well-tested. The two flagged items are edge-case defensive-coding gaps: the missing runs/ existence check in the fresh_summaries comprehension (inconsistent with every other helper in the file) and the lack of a guard that skips the rollup when no fresh summaries are collected. Neither can cause data loss or incorrect findings; the worst outcome is a confusing empty rollup or a version-dependent OSError on Python less than 3.12.

tools/run-sweep.py — the fresh_summaries collection and the unconditional run_sweep_summary call.

Important Files Changed

Filename	Overview
tools/run-sweep.py	Adds aggregate sweep-summary orchestration; removes CODECOME_USE_WRAPPER branch; fresh_summaries collection lacks a runs/ existence guard and doesn't skip the rollup when no fresh summaries are found.
tools/codecome/run_summary_questions.py	Adds exclude_patterns parameter and sweep-exclusion default for phase-2 lookups; adds find_latest_sweep_summary(); implementation is clean and well-tested.
tools/codecome.py	Replaces per-file sweep scan with single aggregate sweep-summary lookup; imports find_latest_sweep_summary; change is clean and correct.
prompts/phase-2-sweep-summary.md	New aggregate rollup prompt; correctly forbids fresh hunting and new finding creation; instructs model to use only the injected per-file summary list.
prompts/phase-2-sweep.md	Renamed from sweep.md; updated artifact naming to phase-2-summary-sweep--*.md; correctly aligned with Phase 2 conventions.
tests/test_run_sweep.py	New test file covering template path, slug, build_prompt_for_file, and build_sweep_summary_prompt; good coverage but no test for the empty fresh_summaries rollup path.
tests/test_run_summary_questions.py	Adds sweep-exclusion and find_latest_sweep_summary tests; comprehensive and well-structured.
tests/test_phases_completion.py	Adds TestSweepCompletionGate verifying the Phase 2 gate accepts sweep-named summaries; tests are correct and cover the key scenarios.

Sequence Diagram

sequenceDiagram
    participant Op as Operator
    participant RS as run-sweep.py
    participant RA as run-agent.py
    participant OC as opencode run
    participant FS as runs/ (filesystem)

    Op->>RS: "make sweep FILE=..."
    RS->>RS: select files, build per-file prompts
    loop Each selected file
        RS->>RA: --phase 2 --prompt-file sweep-slug.md
        RA->>OC: opencode run auditor agent
        OC-->>FS: phase-2-summary-sweep-slug-TIMESTAMP.md
        OC-->>RA: exit code
        RA-->>RS: exit code Phase 2 completion gate checked
    end
    RS->>RS: collect fresh_summaries mtime filter
    RS->>OC: opencode run --agent auditor sweep-summary prompt
    OC-->>FS: sweep-summary-TIMESTAMP.md
    OC-->>RS: exit code
    RS-->>Op: done aggregate rollup printed to screen

    note over Op: make hints
    Op->>RS: codecome hints
    RS->>FS: find_latest_summary 2 excludes sweep files
    RS->>FS: find_latest_sweep_summary returns sweep-summary files
    RS-->>Op: Phase 2 block plus Sweep block no duplicates

Loading

_{Reviews (2): Last reviewed commit: "fix: address PR #55 review comments" | Re-trigger Greptile}

[pruiz]

Thanks for the reviews. All feedback has been addressed in commit 2cc850b:

---

Copilot: docstring says "override" but implementation cannot override

Fixed. Docstring now says "extend" instead of "override". The implicit phase-2-summary-sweep-*.md exclusion for Phase 2 calls is always applied; exclude_patterns can extend but not override it. Also renamed test test_explicit_exclude_pattern_overrides_default → test_explicit_exclude_pattern_extends_default, and switched matching from Path.match() to fnmatch for clarity.

---

Greptile: aggregate summary bypasses CODECOME_USE_WRAPPER

CODECOME_USE_WRAPPER is deprecated (see .project/migrate-to-opencode-serve.md and .project/codeql-integration-plan.md which both call for its removal). Rather than supporting it, the follow-up commit:

• Removed the deprecated CODECOME_USE_WRAPPER branch from run_one_file() entirely — per-file sweeps now always use the wrapper path
• Added a docstring to run_sweep_summary() explaining why the aggregate rollup uses raw opencode run (it is not a phase-mode run and run-agent.py lacks non-phase utility support)
• Cleaned CODECOME_USE_WRAPPER references from README.md and docs/workflow.md

---

Greptile: redundant guards in aggregate-summary call

Fixed. The len(files) >= 1 guard was redundant (earlier if not files: return 0 guarantees non-empty). Simplified to if not args.dry_run: and dropped the dry_run parameter from run_sweep_summary() since it can never be reached during dry-run.

---

Additional fixes in 2cc850b

• Stale summary contamination: main() now records sweep_start_time before per-file runs and collects only fresh per-file summaries (mtime >= sweep_start_time) matching phase-2-summary-sweep-{slugify(file)}-*.md. These exact paths are injected into the generated aggregate prompt.
• Aggregate prompt updated: now tells the model to read only the injected per-file summary paths, not all historical phase-2-summary-sweep-*.md.
• Slug example fixed: upload-php → upload.php in the per-file sweep prompt to match slugify() output.
• Timestamped aggregate prompt path: tmp/file-sweep-prompts/sweep-summary-YYYYMMDD-HHMMSS.md instead of fixed sweep-summary-prompt.md.

All 825 non-MockLLM tests pass + frontmatter validation ✓

[pruiz]

Replying to review threads: all feedback addressed in 2cc850b.

[pruiz]

All review feedback addressed in 2cc850b. Thank you both for the thorough review.

[pruiz]

@greptile-apps: your summary analysis and the two inline comments on tools/run-sweep.py were addressed in commit 2cc850b:

• CODECOME_USE_WRAPPER inconsistency: This env var is deprecated project-wide. Removed the legacy branch from run_one_file() entirely. Aggregate rollup uses raw opencode run by design (not a phase-mode run; documented in the function docstring). Cleaned CODECOME_USE_WRAPPER from README.md and docs/workflow.md.

• Redundant guards: Simplified to if not args.dry_run: and dropped the dead dry_run parameter from run_sweep_summary().

• Additional fix: The aggregate prompt now injects exact per-file summary paths (filtered by sweep_start_time) instead of telling the model to read all historical phase-2-summary-sweep-*.md.

Thank you for the review.

[pruiz]

@copilot-pull-request-reviewer: your docstring concern on tools/codecome/run_summary_questions.py was addressed in commit 2cc850b. The docstring now says "extend" instead of "override", matching the implementation where phase-2-summary-sweep-*.md is always excluded for Phase 2 calls regardless of exclude_patterns. The corresponding test was renamed and matching switched to fnmatch for clarity. Thank you.

[greptile-apps]

All the changes look correct and address the feedback precisely:

• CODECOME_USE_WRAPPER branch removed from run_one_file() — per-file runs always use run-agent.py
• run_sweep_summary() has no dry_run parameter; the guard at the call site is a clean if not args.dry_run:
• sweep_start_time = time.time() is captured before the loop, and fresh_summaries is filtered by st_mtime >= sweep_start_time
• Aggregate prompt gets exact timestamped paths injected, with a clear instruction to read only those summaries
• The docstring on run_sweep_summary correctly explains the raw opencode run choice
• Prompt path is now timestamped (sweep-summary-YYYYMMDD-HHMMSS.md)

Everything checks out. The stale-summary filtering via sweep_start_time is a particularly solid fix — it's both correct and robust to partial reruns. Thanks for the thorough follow-up.

Tip: You can customize Greptile's behavior for this repo with .greptile/rules.md and .greptile/config.json.