gaofeng21cn

Design domain-neutral Foundry EvalSpec cases, gates, protected-test requirement categories, failure taxonomies, and EvidenceBundle expectations. Use when creating or reviewing evaluation semantics for an AgentBlueprint without taking over protected tests, execution, qualification, or Owner acceptance.

Interpret OPL FoundryRun, EvidenceBundle, qualification, risk, Owner-gate, canary, activation, and rollback evidence. Use when an operator needs an evidence-bound failure classification or next route for an Agent create, takeover, or improve run.

Design OPL Stagecraft support for stage goal, prompt shaping, rubric, capability-use decisions, route-back reasoning, handoff, and receipt/blocker lower-bound design. Use when authoring or reviewing an OPL stage strategy or stage prompt wrapper while keeping Stagecraft contracts limited to descriptors, allowed/forbidden surfaces, and no-authority receipt/blocker lower bounds.

Use when reviewing OPL family stage assumption lifecycle projections, stale or missing assumptions, monitor gaps, owner gaps, minimal counterexamples, repair actions, and no-authority route-back recommendations.

Use when reviewing OPL stage candidate portfolio refs-only projections, candidate ref coverage, assumption/provenance/negative-path/advisory-metric refs, human-review route-back, missing refs, and domain-authority overclaim risks.

Use when reviewing OPL family stage-pack source specs and replay certifications, including body-free source diffs, visual-equivalent spec gaps, replay blockers, missing runtime event refs, missing receipt refs, and owner-route workorders.

Use when critiquing OPL Stagecraft quality gates, stage admission projections, stage output evidence, evaluator framing, rubric gaps, trust lanes, composition obligations, failure localization, human-review burden, owner-route coverage, route-back packets, and no-authority gate recommendations.

Use when writing or reviewing OPL Workspace handoff packets, source-readiness audit sections, source refs, artifact refs, workspace briefings, missing-input route-back notes, and owner-route packets while keeping locator, path, artifact-unit, and source-readiness truth in Workspace or program-owned surfaces.

Use when a quest has enough evidence to draft or refine a paper, report, or research summary without inventing missing support.

Use when the quest needs an explicit go, stop, branch, reuse-baseline, write, finalize, reset, or user-decision transition with reasons and evidence.

Use when the quest is ready to consolidate final claims, limitations, recommendations, summary state, and graph exports before stopping or archiving.

Use when a quest needs one or more follow-up runs such as ablations, robustness checks, error analysis, or failure analysis after a main experiment.

Use when a quest is ready for a concrete implementation pass or a main experiment run tied to a selected idea and an accepted baseline.

Use when a quest needs concrete hypotheses, limitation analysis, candidate directions, or a selected idea relative to the active baseline.

Use when a draft, paper, or paper-like report is substantial enough for an independent skeptical audit before finalization, rebuttal, or revision routing.

Use when a quest needs to attach, import, reproduce, repair, verify, compare, or publish a baseline and its metrics.

Use when Codex needs RedCube AI (RCA) to create, revise, review, or package a visual deliverable such as a slide deck/PPT, poster, social visual, or visual handoff. Ordinary PPT/PPTX requests default to full-slide images packaged in a .pptx container; select native editable PowerPoint only when the current user explicitly requires editable/native objects. Do not use for Agent engineering merely because OMA is mentioned.

Use when Codex needs RedCube AI (RCA) to create, revise, review, or package a visual deliverable such as a slide deck/PPT, poster, social visual, or visual handoff. Ordinary PPT/PPTX requests default to full-slide images packaged in a .pptx container; select native editable PowerPoint only when the current user explicitly requires editable/native objects. Do not use for Agent engineering merely because OMA is mentioned.

Use when RedCube AI needs a native editable PowerPoint specialist to produce AI-authored editable_shape_plan, design spec lock, template grammar, shape coordinates, and native PPTX repair targets under RCA native route rules.

Use when RedCube AI needs a PPT visual direction specialist to convert an approved slide blueprint into visual language, rhythm, density controls, layout variety, and anti-template constraints before artifact creation.

Use when RedCube AI needs a PPT page authoring specialist to create or repair individual slide pages from approved story and visual direction while preserving RCA route policy and review gates.

Use when RedCube AI needs a PPT review specialist to inspect rendered deck pages, screenshots, source fidelity, layout quality, and repair targets before export or handoff.

Use when RedCube AI needs a PPT story specialist to turn frozen source truth into a claim spine, narrative sequence, detailed outline, or slide blueprint for an RCA ppt_deck stage without producing visual artifacts.

Use when RedCube AI needs Xiaohongshu single-note or series content strategy grounded in source truth, audience needs, platform constraints, and cross-note continuity.

Assess an existing Agent version and design a frozen, independent evaluation specification.

Use only when Codex is explicitly asked to create, take over, assess for evolution, or improve an OPL-compatible Agent, including its architecture, prompts, skills, contracts, evaluations, or authority boundaries. Do not use because OMA was mentioned or @-mentioned, another Agent is producing a deliverable, or a deliverable validator, render, or QA check failed.

Admit design evidence and turn an Agent mission into a coherent Stage and authority architecture.

Diagnose an exact OPL EvidenceBundle without taking evaluator or runtime authority.

Produce an evidence-bound complete next AgentBlueprint and semantic change set.

Use when Med Auto Grant needs proposal authoring, independent review, rebuttal planning, or local submission-package audit after strategy and aims are accepted.

Use when Med Auto Grant needs call-fit, fundability, specific-aims, or strategy-memory judgment before proposal authoring.

Use when Codex needs Med Auto Grant (MAG) to plan, author, critique, revise, or package a medical grant application, including funding-call intake, fundability strategy, specific aims, proposal sections, rebuttal, and a submission-ready local package. Do not use for research-paper production, generic document formatting, patient care, or irreversible submission to a sponsor portal.

Use when Codex needs Med Auto Grant (MAG) to plan, author, critique, revise, or package a medical grant application, including funding-call intake, fundability strategy, specific aims, proposal sections, rebuttal, and a submission-ready local package. Do not use for research-paper production, generic document formatting, patient care, or irreversible submission to a sponsor portal.

Scheduled task management - create, query, update scheduled tasks to automatically execute operations at specified times.

Use when tasks involve reading, creating, or reviewing PDF files where rendering and layout matter; prefer visual checks by rendering pages and use Python tools such as reportlab, pdfplumber, and pypdf for generation and extraction.

Use when extracting content from PDFs, scanned documents, Office files, images, or web pages into Markdown, HTML, LaTeX, DOCX, or JSON, especially when OCR, tables, formulas, or batch document conversion are required.

Use when inspecting or upgrading managed local Codex skills against explicit upstream sources, or when syncing a Skills Manager central library across machines without re-deriving source mappings by hand.

Use when turning academic or medical-research materials into defense deliverables, including timed oral scripts, PPT speaker notes, narrative/storyline rewrites, reviewer-facing questions with background notes, or paired slide-notes plus script outputs. Trigger on PPT/PPTX plus source PDFs, thesis zip packages, proposal/application materials, dissertation defenses, project defenses, or requests such as "5-minute script", "speaker notes", "defense questions", "paced oral defense", or "storyline". Pair with Office/PPTX/PDF extraction skills for file operations.

Use when managing Apple Mail on macOS, especially to list accounts or mailboxes, inspect recent messages, search, read, mark, archive, move, or delete emails in Mail.app; also use it as the shared entry point for future Apple native app automation on this Mac.

Use when Codex needs MedAutoScience (MAS) to plan, conduct, review, or publish medical research, including research questions, study design, cohorts and endpoints, evidence synthesis, bounded analysis, scientific figures/tables, manuscript writing, quality review, and publication handoff. Do not use for patient-specific diagnosis, treatment, triage, or emergency advice; medical grant applications; or generic Office formatting without a research objective.

Use when Codex needs MedAutoScience (MAS) to plan, conduct, review, or publish medical research, including research questions, study design, cohorts and endpoints, evidence synthesis, bounded analysis, scientific figures/tables, manuscript writing, quality review, and publication handoff. Do not use for patient-specific diagnosis, treatment, triage, or emergency advice; medical grant applications; or generic Office formatting without a research objective.

Testing workflow and quality standards for writing and running tests. Use when: (1) Writing new tests, (2) Adding a new feature that needs tests, (3) Modifying logic that has existing tests, (4) Before claiming a task is complete.