Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Render Lineage Auditor

    1

    Reconcile render specs, mission/run/job lineage, artifact hashes, CAS copies and release receipts—without treating mock or local success as distributed proof.

    Free

    Security scanned
    Render Lineage Auditor

    Render Lineage Auditor

    Example session with this skill installed

    Use Render Lineage Auditor on the bundled demo only. Run the audit against examples/sample.mp4, examples/spec.json, examples/evidence.json, and examples/sample.cas.mp4. Show the measured media properties, calculate and compare the full SHA-256 hashes, reconcile delivery and verification/release records, and explain whether the supplied evidence proves a real remote render. Do not modify any input files. Treat every bundled identifier and remote-worker field as synthetic demo data, not production evidence.

    • Read your context and instructions
    • Compiled the render lineage auditor

    Summary

    The media measurements and local CAS byte comparison passed, but the overall verdict is

    INSUFFICIENT EVIDENCE. This is the expected result for the bundled synthetic fixture: its metadata claims remote execution, but the tool does not authenticate a cloud worker from a JSON field.

    Measured artifact

    CheckMeasured resultResult
    Duration5.000 secondsPASS (target 5 s ± 0.25 s)
    VideoH.264, 360 × 640PASS
    Frame rate30/1 fpsPASS
    AudioAAC stream presentPASS
    File SHA-2566034acc6f13bef9490c40a11e15e98ddc0dbc54619a58afde1e80ee6662aec7dCalculated from actual bytes
    CAS copySame full SHA-256 as the media filePASS — byte equality only

    Execution and release evidence

    Evidence itemObservationInterpretation
    Mission/run/job IDsPresent in the demo JSONIDs are synthetic and not authenticated
    Execution moderemote with demo provider/worker IDsClaim recorded, not authenticated
    Delivery recordUPLOADED, job ID and hash agree with the sampleConsistent fields, source not authenticated
    Physical-verification recordVERIFIED / PHYSICAL_VERIFIER; hash and lineage fields agreeConsistent fields, receipt authenticity not established
    Release recordVERIFIED; hash and lineage fields agreeConsistent fields, receipt authenticity not established

    Verdict: INSUFFICIENT EVIDENCE

    There is no measured media mismatch in this fixture, and the two local files are byte-identical. However, the accompanying job, worker, delivery and release identifiers are deliberately synthetic.

    This example therefore does

    not demonstrate a real remote render, authenticated worker identity, live production delivery, or distributed storage durability.

    Next steps

    1. Replace the demo fixture with the real output file and its original, unmodified job/mission/run and release records.
    2. Compare the recorded artifact hash and the actual CAS object hash.
    3. If remote execution is required, obtain separate trusted worker-verification evidence; don't rely only on a remote field.
    4. Keep any missing or unauthenticated step as INSUFFICIENT EVIDENCE and report the smallest gap to close.

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

    About this skill

    Know whether a completed render is actually supported by evidence

    A pipeline can report success while returning the wrong file, pointing at a stale CAS object, losing mission/run/job lineage, or presenting a local/mock result as a distributed render. Render Lineage Auditor checks the relationship between the requested media profile, the real file bytes, the supplied execution records and the delivery/release evidence.

    What it checks

    • Measures actual video codec, dimensions, frame rate, duration, and audio stream/codec with ffprobe when those properties are specified.
    • Calculates SHA-256 from the actual artifact bytes and compares any recorded artifact hash.
    • Hashes an optional local CAS copy and checks whether its bytes exactly match the delivered artifact.
    • Reconciles mission, run and job identifiers across supplied records.
    • Checks delivery status, physical-verification records, and release receipts for consistent status, hash and available lineage fields.
    • Produces a structured JSON report with explicit PASS, PASS_WITH_SCOPE_LIMITS, FAIL, INSUFFICIENT_EVIDENCE, or BLOCKED verdicts.

    What makes it different

    This is a pipeline provenance and evidence-consistency gate, not a deep media-quality suite. It asks whether the file, expected spec, CAS object and pipeline receipts all line up.

    A local CAS hash match proves byte equality only; it does not prove shared storage or cross-instance durability. A JSON field saying remote or VERIFIED remains a supplied claim unless its source is independently authenticated.

    Safe by default

    The included script does not make network requests, read environment variables, transcode media, or modify input files. It uses Python's standard library and ffprobe; it writes a report only when an output path is requested and refuses to overwrite an existing report unless you explicitly pass --overwrite.

    Requirements

    Python 3.9+ and ffprobe on PATH (normally installed with FFmpeg). No API keys, browser automation, cloud account, or third-party Python package is required for the local audit.

    Limitations

    It does not authenticate cloud-worker identity or cryptographic signatures, prove remote object-store durability, judge visual scene quality, perform frame-by-frame editing checks, or measure detailed audio loudness. It reports only what it can measure and what supplied records support. Use specialized media-QA tools for deeper codec, loudness, color, frame and visual-quality checks.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.

    Reads the selected artifact and supplied JSON records. Performs no network requests and reads no environment variables. Does not modify the source media or receipts. Writes only the requested report path; an existing output file is preserved unless --overwrite is explicitly passed.
    Known limitations

    • Requires a local readable file and ffprobe for measured media properties; does not fetch URLs.
    • Does not authenticate remote worker identity, provider claims, or cryptographic signatures by itself.
    • Local CAS equality does not prove distributed durability or cross-instance access.
    • Does not replace detailed visual-quality, frame-by-frame, color, audio loudness or silence analysis.
    • Generic evidence schemas may need a temporary normalized copy for a specific platform; original records should remain untouched.

    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 buying a skill.

    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

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 23 days ago

    • Free with an Agensi account

    Software
    Python 3.9+ffprobe (normally installed with FFmpeg)

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    **/*.mp4
    **/*.mov
    **/*.mkv
    **/*.webm
    **/*.gif
    **/*.json
    **/render-lineage-report*.json
    Listed23 days ago

    What's inside

    Frequently Asked Questions

    No. It checks media facts and consistency of supplied records. A remote field is not authentication. Use trusted, separate worker verification if remote execution must be proven.

    The skill requires a Python 3.9+ environment with ffprobe installed on the system path, and it must have read-access to the local directories where your media artifacts and JSON job records are stored.

    Any agent that can read a standard SKILL.md directory and run terminal commands. The included Python script was tested in the development environment; every agent/OS pairing has not been independently tested.

    You will receive the proprietary auditing logic, the bundled Python verification script, and clear integration documentation to connect the skill to your agent's workflow.

    It is compatible with any AI agent or framework capable of executing shell commands and reading local files, making it ideal for LLM-ops and automated video production pipelines.

    Unlike standard prompts that rely on surface-level logs, this skill uses a 'trust but verify' architecture that reports INSUFFICIENT_EVIDENCE if a hash or lineage record is missing, ensuring zero-trust artifact validation.

    New in DevOps & Cloud

    Free