dKosarevsky/albu-mcp / skills/albumentationsx-mcp
albumentationsx-mcp
VerifiedUse when Codex needs to install, configure, or operate AlbumentationsX MCP from agent hosts for computer-vision augmentation previews, feedback-driven tuning, bounded local dataset review, or evidence-backed adoption workflows.
AlbumentationsX MCP
Install
uvx --from albumentationsx-mcp albumentationsx-mcp
uvx --from albumentationsx-mcp albumentationsx-mcp \
--allowed-root /absolute/path/to/images \
--artifact-root /absolute/path/to/albu-artifacts
First Run Prompt
Use this as the first host task:
Use AlbumentationsX MCP on image or directory `DATASET_PATH`. Read from `ALLOWED_ROOT` and write to `ARTIFACT_ROOT`. When the host exposes resource reads, read `albumentationsx://examples/client-smoke`; if resource reads are unavailable, call `get_workflow_example` with `example_id="client-smoke"`. Then call `run_host_smoke_check`. Continue only when `preview_ready` is true. If `preview_ready` is false, call `diagnose_environment` and stop before rendering. Then call `plan_dataset_onboarding`, `build_review_packet`, `validate_preview_request`, and `render_preview_batch`. Render at most 6 images on the first pass. Show the contact sheet path and ask for concrete feedback before `adjust_pipeline` or `export_pipeline`.
Host Config Hints
- Codex plugin mode uses
.codex-plugin/plugin.jsonand.mcp.json; its pinned server grants no user dataset root. - Set
ALBU_MCP_ALLOWED_ROOTSandALBU_MCP_ARTIFACT_ROOT, or use explicit absolute host args. Never rely on the working directory. - Restart, run
run_host_smoke_check, and stop unlessallowed_rootscontains the intended root andpreview_readyis true.
Host Workflow
- Read
albumentationsx://examples/client-smoke; if resource reads are unavailable, callget_workflow_examplewithexample_id="client-smoke". - Call
run_host_smoke_checknext; continue only whenpreview_readyis true. - Call
plan_dataset_onboarding, thenbuild_review_packetfor one image or folder. - Validate user paths with
validate_preview_requestbefore rendering. - Render a small sample with
render_preview_batch; inspect the contact sheet. - Record feedback with
record_preview_feedback, such astoo_noisy:highorexposure_too_weak. adjust_pipeline, re-render, and compare before acceptance.- Export only reviewed work with
export_pipelineor requested report tools.
Safety Rules
- Do not train, overwrite datasets, or fetch remote images.
- Keep private paths, filenames, and image contents out of public reports.
- Generated fixtures, contact sheets, and rehearsals are not beta evidence.
- Read only under
--allowed-root; write only under--artifact-root. - Re-run validation after changing paths, masks, bboxes, labels, or annotation formats.
Stop Conditions
- Missing real image or dataset-directory path: ask for one.
- Path outside
--allowed-root: refuse that path and ask for a bounded path. - User asks for many variants: render a small first batch before expanding.
Evidence Workflows
For evidence work, follow the generated pack README and use:
albu-mcp activation real-adoption-cyclealbu-mcp activation product-fix-closure-pipelinealbu-mcp evidence execution-pack --date YYYY-MM-DD --reviewer "Release operator" --output-dir evidence-session --format markdownalbu-mcp evidence execution-pack-audit --input-dir evidence-sessionalbu-mcp evidence execution-pack-progress --input-dir evidence-sessionalbu-mcp evidence execution-pack-status --input-dir evidence-session --format markdown --output evidence-session/status.mdalbu-mcp evidence preflightalbu-mcp evidence import-wizard
If preview setup fails, read albumentationsx://diagnostics/guide, call diagnose_environment, then remediate before rendering.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |