review-consolidate
Run repository/lens reviews in separate processes, then consolidate validated, scope-aware full reports.
Skill by gbencke
Chalk Labs
Need help deploying review-consolidate?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/gbencke-review-consolidate-9492a8
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/review-consolidate
What's inside
- Skill name
review-consolidate- Description
- Run repository/lens reviews in separate processes, then consolidate validated, scope-aware full reports. Focused verifications and legacy reports remain history and cannot replace full-review status.
- License
- MIT
- Instructions
- 104 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# review-consolidate Read `docs/report-contract.md` from the extension directory. Consolidation arranges existing evidence; it does not review source, invent findings, or reclassify severity. No subagents. Parallelism is the existing matrix launcher. ## 0 — Arguments Require at least one repo and `--out`; never default either. Resolve paths to absolute and permit a directory prefix only when exactly one repo matches. `--lenses` limits the matrix, `--no-run` reads existing artifacts, and `--all` includes historical full snapshots in an appendix, **not** a union of old and current defect counts. Stop with usage when arguments are missing/ambiguous. Use a UTC stamp and a sibling `<out-directory>/review-matrix-<stamp>.logs` directory. ## 1 — Run once, unless --no-run ```bash <extension>/bin/run-matrix.sh --log-dir <logdir> [--lenses a,b,c] <repos...> ``` One process per repo/lens, all launched before waiting. Use a generous timeout; never restart the launcher, spawn subagents, or silently retry failed pairings. Read `manifest.tsv`: repo, lens, exit, seconds, report, log. The launcher validates v2 sidecars before accepting a REPORT path; missing/invalid reports are failures. A failed pairing stays FAILED for this run. Do **not** substitute an older success and make current coverage appear complete. Historical artifacts may be linked separately, explicitly dated. ## 2 — Collect by identity, revision, and scope For successful manifest pairings, use exactly their reported paths and sidecars. Validate structural/history consistency; do not infer a newer report from mtimes. For `--no-run` use: ```bash node <extension>/bin/report-tools.mjs select <absolute-repo> [--all] ``` The selector groups by target/PR, mode, and explicit lens set, then orders parsed UTC timestamps. It selects only **full v2** reports. Its `ignored` list includes legacy/focused artifacts and reasons: preserve that list in the output. Do not let a recent `verification`, a focused `multi`, or a legacy zero-count report replace a full review. These artifacts remain visible as history without a new whole-repository/PR approval. The latest full report already contains its cumulative ledger. A later focused verification may settle its named issue, but does not silently recalculate that full report's verdict: list the verification separately, and require a reconciled full report before presenting a new approval. Do not resurrect old issues by unioning every historical snapshot's findings. If parsing/validation fails, record the exact artifact and error as UNPARSEABLE / INVALID. If there is no full v2 report, record NO VALID FULL REPORT, even if legacy history exists. Do not claim an empty repository is healthy. Report: repos, pairings run, full sidecars selected, active confirmed findings, verification questions, historical/focused artifacts, and failed pairings. ## 3 — Arrange the evidence 1. Keep target, head SHA, mode, scope, and lens identity on every row. Reports of different revisions are not equivalent current checks. Flag mixed-head runs; never present them as one current-revision certification. 2. Within the same repo/target/head, merge repeated stable IDs/explicit aliases, preserving evidence and lenses. If two reports conflict about disposition, retain both observations and mark reconciliation required; do not adjudicate. 3. Active means `status: open`, or `not_rechecked` with `prior_status: open`. Count VALID separately from AMBIGUOUS. Fixed, disproved, and deferred records belong in a disposition section, not the active-defect totals. 4. Never merge findings across repositories. Shared defect classes may receive CROSS-REPO labels, but each owner retains its own finding and evidence. 5. Preserve judged severity. Multiple labels/lenses are not independent evidence. ## 4 — Write --out Create its parent; write only the requested output, not inside reviewed repos unless that is the explicit output path. Do not modify their source or reports. Structure: - Summary: repo/target/revision/scope, confirmed P0–P3, uncertain P0–P3, verdict. - Coverage: each requested pairing, report date/head, validation status and log. - Active confirmed findings: severity, then repo, with failure/evidence/fix. - Verification questions: uncertainty and the exact check needed. - Finding dispositions: verified fixes, disproofs, deferrals, and unrechecked issues. - Cross-repository themes, if supported. - History: focused/legacy reports and optional `--all` snapshots, clearly non-current. - Not consolidated: failed/missing/invalid reports, conflicting dispositions, mixed revisions, and incomplete coverage. Never hide these behind zero counts. A failed or incomplete pairing prevents an overall healthy/SHIP summary. Do not rescue an INCOMPLETE report by combining unrelated partial outputs. Read-only artifact collection is the only work here; no source review or application runs. ## 5 — Summary Return counts with scope/revision qualifications, pairings run/failed, top five confirmed findings, outstanding verification questions, and output path. Last line: `CONSOLIDATED: <absolute-output-path>`.
From gbencke/dotfiles, MIT.
Signals
- Stars
- 6
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 30, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/gbencke/dotfiles
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get review-consolidate running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
review-consolidate skill. Agent Mart. https://agentmart.dev/skills/gbencke-review-consolidate-9492a8 (last updated 2026-10-24)