PromptBP Repository & PR Review

    1

    Get prioritized PR or repository findings with source evidence, coverage gaps and a review receipt.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    PromptBP Repository & PR Review

    Example session with this skill installed

    Review dburt-proex/PromptBP PR #11 at base 95f212b37b005db8b23bc0cf9109b27481565736 and head 729d82fbdbf0692f368d3cc1b68b9c68bdd6e8d9 for actionable functional and delivery defects. Keep the target read-only. Account for all changed files, inspect validation effects before running anything, preserve existing changes, and return findings, coverage, checks, limitations and a bound receipt. No external writes, installs or setting changes.

    • Read your context and instructions
    • Compiled the promptbp repository

    Recorded creator evaluation — October 22, 2026

    This is the complete human-readable output from the independent creator-local evaluation, followed by its actual receipt. It is a review of the old 0.2.0 PR candidates, not a paid buyer transaction or a claim that those candidates were release-ready. The final 0.3.0 instruction text was subsequently verified through Agensi creator-test delivery. Evidence filenames refer to the retained creator evidence packet.

    Review of dburt-proex/PromptBP PR #11 at head 729d82fbdbf0692f368d3cc1b68b9c68bdd6e8d9 (base 95f212b37b005db8b23bc0cf9109b27481565736).

    Two P2 defects reproduced in the pinned 0.2.0 distributions:

    1. A one-repository review directive and environment copy the full operator scope, including unrelated repositories. Narrow both emitted scopes to the requested target. Evidence: scope-counterexample.json; bundled promptbp/spine.py lines 222–225 and environment.py lines 121–122.
    2. The documented preparation helper discards the buyer's bounded objective and acceptance criteria by hard-coding a generic repository-state request. The receipt can report generic completion without the original objective being encoded. Accept a validated intent/criteria input. Evidence: objective-counterexample.json; bundled prepare_review.py lines 35–43.

    These behaviors are inherited upstream code newly distributed by this PR, not changes to the root runtime. Both bundles and ZIPs contain them. No target fixes were attempted.

    All 84 changed paths are accounted for in coverage.json. Each pack passed 54 bundled tests plus demo, sample validation and starter compilation using existing Python 3.11.15 / PyYAML 6.0.3. Tests ran only in the fresh pinned scratch copy after side-effect inspection. ZIP CRC/safe paths and all 27 member text comparisons passed; 22 runtime files per pack equal pinned root files. Local git export converted text to CRLF, while ZIPs retain LF, so archive equality is normalized text equality, not raw-byte equality.

    GitHub recheck confirms closed/unmerged draft PR and unchanged head. Sample validation CI passed; DiffWall remains HALT 80/100. No submitted reviews or threads returned. Historical blocked receipt recomputed exactly. Native installation, marketplace security scan and successful live execution of the pinned 0.2.0 workflow remain unverified.

    Source preservation passed: all 141 tracked/nonignored file hashes, HEAD and git status are unchanged, including the three existing user modifications. Git internal metadata and ignored files were excluded.

    Review status: reported_completed within the recorded coverage. Code/release status: not ready. Read-only authority gate: ALLOW; no merge, publication, external write, installation or settings change authorized or performed.

    Next action: correct the two handoff defects in a separately authorized candidate and retest it before release.

    Skill forward-test: creator-local 0.3.0 workflow worked with unknown model metadata. Its Agensi-load instruction was overridden by the explicit local forward-test request. Prose-only result/receipt formats required manual artifact assembly. Larger combined reads truncated output; relevant sections were re-read. Full result, checks, limitations and evidence pointers are in result.json; SHA binding is in receipt.json.

    Actual bound receipt

    {
      "plan_id": "independent-pr11-20261022-01",
      "binding": "SHA-256 of exact saved output bytes",
      "plan_sha256": "a3631642d2cef28078707bdb8884f1b9289730a871a6e2e2167d86973c80cf77",
      "result_sha256": "c899f8d04cf15733902a6a4dd3f340155b18ae0575ffdcea4f57019a2b8a70c1",
      "readable_report_sha256": "4019c79e7106116728058b6cc154c5174df6aa8e8466e03830bd9d0aecca6833",
      "target": "dburt-proex/PromptBP#11",
      "base_sha": "95f212b37b005db8b23bc0cf9109b27481565736",
      "head_sha": "729d82fbdbf0692f368d3cc1b68b9c68bdd6e8d9",
      "reported_by": "Codex independent review subagent /root/review_forward_test",
      "reported_at": "2026-10-22T07:49:39.5436331Z",
      "host_metadata": {
        "thinking": "unknown",
        "model": "unknown",
        "configuration_changes": []
      },
      "acceptance_checks": [
        {
          "status": "passed",
          "name": "Exact target and revisions verified"
        },
        {
          "status": "passed",
          "name": "Coverage accounted for including gaps"
        },
        {
          "status": "passed",
          "name": "Evidence-backed findings"
        },
        {
          "status": "passed",
          "name": "Optional checks accurately labeled"
        },
        {
          "status": "passed",
          "name": "Source preservation confirmed"
        }
      ],
      "evidence_references": [
        "plan.json",
        "result.json",
        "coverage.json",
        "source-before.json",
        "source-after.json",
        "source-preservation.json",
        "pack-checks.json",
        "scope-counterexample.json",
        "objective-counterexample.json",
        "historical-evidence-audit.json",
        "github-initial.json",
        "github-closeout.json",
        "diffwall-log-excerpt.txt"
      ],
      "execution_status": "reported_completed",
      "limitations": "Executor-reported receipt; no independent evidence certification or release approval. CI HALT and identified defects remain."
    }
    
    scope-counterexample.json

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Give your agent one repository or PR and a clear review objective. PromptBP establishes exact scope, checks available read tools, reviews code and tests, and returns actionable findings with file, line and revision evidence. The PR workflow pins base/head commits, accounts for every changed file, checks the head again at closeout, and marks stale or incomplete coverage explicitly. Supplied diffs without live revision proof receive an explicitly limited supplied-evidence review.

    You get a bounded review plan, prioritized findings or a scoped no-findings result, an honest test/coverage record and an executor-reported receipt. With an authorized host output/hash tool, the receipt binds the exact plan and result bytes using SHA-256. Without one, it explicitly reports binding unavailable. A successful review can find defects; it is not a merge approval or a claim of defect-free software.

    Required: authorized repository/file read tools or supplied source evidence. Local live review needs local read tools. Saving/hashing results needs a capable host tool and permission for an evidence folder outside the target. No fixes, installs, GitHub review submissions, merge, deployment or settings changes are performed. Optional tests run only after side-effect inspection and within scope.

    Recorded creator demonstration, October 22, 2026: a fresh Codex evaluator manually loaded this workflow and reviewed PromptBP PR #11 at head 729d82f. It accounted for all 84 changed paths, independently reproduced two P2 handoff defects in the old Python packs, preserved 141 source-file hashes and produced a SHA-256-bound reported_completed review receipt. The PR itself remained closed/unmerged with a DiffWall HALT; review completion did not approve release. This is observed local workflow evidence, not an Agensi installation, marketplace security scan or universal compatibility benchmark.

    Agent-native edition 0.3.0. Loads through the Agensi connection without package extraction, local skill installation, Python or a companion purchase. Uses existing host tools; no credential collection, automatic model switching, tool activation or external writes. Host permissions and billing remain the user's responsibility. Exact model/thinking metadata remain unknown when unexposed; explicit configuration requirements must be verified. Receipts are executor-reported; hashes detect byte changes, not truth or signatures. This edition does not invoke or claim deterministic validation from the separate PromptBP Python compiler.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 26 days ago

    Permissions

    No special permissions declared or detected

    Listed26 days ago

    What's inside

    Frequently Asked Questions

    Get prioritized PR or repository findings with source evidence, coverage gaps and a review receipt.

    This skill uses the open SKILL.md standard. Use it through an MCP-compatible agent connected to Agensi, and check the compatibility information on this page for confirmed agents and required capabilities.

    Connect your AI agent to Agensi through MCP, then ask it to use PromptBP Repository & PR Review on your task. Your agent loads the latest approved version each time. Paid skills require a purchase or active subscription. Skill files are not available for download; your licence does not allow you to save, copy or export any part of a skill.

    New in Code Quality & Review