菜单
Decide which research project skills to use for each task, whether the task should be handled by one agent or split across multiple subagents, and how to coordinate outputs safely.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Design and review active learning implementation support for university teachers, including classroom activity structures, preparation workflow, student participation, assessment alignment, and AI-agent support points.
Turn dissertation evidence into an AI agent design specification for supporting active learning implementation, including user scenarios, functions, interaction flow, boundaries, failure modes, and design rationale.
Synthesize confirmed co-design, design-elicitation, prototype-feedback, or participant-generated design outputs into requirements, design principles, prototype changes, participant contribution maps, and design decision logs.
Design specifications for dissertation figures such as conceptual frameworks, concept-card/interview process diagrams, AI-agent workflow diagrams, adoption-condition models, and findings maps.
Maintain a lightweight dissertation research wiki with project memory, literature notes, methodology decisions, supervisor feedback, design rationale, and open questions.
Audit an AI agent prototype or design concept for usability, pedagogical fit, teacher feedback, active learning support, evidence quality, and next design iterations.
| name | agent-orchestration |
| description | Decide which research project skills to use for each task, whether the task should be handled by one agent or split across multiple subagents, and how to coordinate outputs safely. |
Use this skill at the start of non-trivial research tasks, especially when the user asks for planning, reviewing, literature work, source ingestion, formal drafting, compliance/ethics review, LMS/portal analysis, or multi-part project updates.
First classify the task. Then choose skills. Then decide whether subagents are useful and allowed.
For the Production Window, this is automatic for every user task. Do not wait for the user to name a skill. For simple chat answers, route silently and answer directly. For substantial research work, formal drafting, source ingestion, Word output, or stakeholder-facing material, state the selected skills briefly before acting and record the routing in the Production run receipt.
Use research-project-adapter when the project type is unclear or when adapting this starter kit to a non-dissertation profile.
The user may grant project-level standing permission for the agent to decide whether subagents are useful. This permission does not mean subagents should be used automatically. Use them only when they materially improve quality, speed, independent review, or parallel research. If the work is narrow, sensitive, or a single-document edit, handle it locally.
Before drafting any formal document, use dissertation-source-first-gate. Extract known factual fields from source files. Do not invent names, emails, dates, institutional contacts, supervisor/PI/client details, signatures, participant facts, datasets, results, administrative requirements, journal/funder/client requirements, marking criteria, grade-band standards, LMS requirements, deadlines, word counts, or submission rules. If a field is not found in the source material, mark it TO CONFIRM.
Before substantive cross-stage work, apply research-wiki/STAGE_CONTINUITY_PROTOCOL.md when both conditions hold: (A) the task changes, designs, restates, translates, formalises, or produces a high-risk deliverable class such as method plan, compliance material, fieldwork instrument, analysis plan, stakeholder-facing decision memo, or formal section draft; and (B) research-wiki/STAGE_GRAPH.md lists upstream dependencies or the task references prior decisions, accepted outputs, source-of-record files, source maps, or checkpoints. Citation-only edits, formatting/layout fixes, typo repair, file moves, bookkeeping, and source summaries without design/formal-claim output do not trigger the gate.
Use scripts/stage_recall_policy.py or the runtime preflight recall_decision as the token-aware recall controller. Tier 0 means no project recall; Tier 1 anchor scan; Tier 2 pointer lookup; Tier 3 targeted Stage Continuity Capsule; Tier 4 full upstream audit or pause. This controller manages context budget only. It never overrides Stage Continuity A+B, source-first, compliance, citation, privacy, document-quality, or delivery gates. If the user says to skip upstream checks for a triggered deliverable, surface the omitted dependency first and record an override risk only after explicit user acceptance.
Keep bounded work bounded. Source planning, reading-priority sorting, source lookup, and minor citation/typo edits should use the light runtime receipt set unless the user asks for formal prose, Word/DOCX, stakeholder-facing or submission-facing output, or protected source-of-record edits. Do not route into the full formal-writing chain merely because the task mentions methodology, literature review, proposal background, or sources.
Before answering rubric, grade-band, marking-criteria, journal/funder/client requirement, LMS requirement, word-count, deadline, or submission-rule questions, use the strongest available project requirement source. Use university-guidance/RUBRIC_EVIDENCE_GATE.md for assessed academic work. Distinguish official original text, local summary, inference, and evidence-insufficient status.
Before delivering formal documents or important project notes, use dissertation-document-quality-gate at the appropriate level.
For major proposal, manuscript, report, grant, methodology, literature review, or stakeholder-facing writing, use source-first, material-passport, and academic-integrity-preflight before cognitive-frameworks. The cognitive output should name section type, claim, gap/problem type, evidence, warrant, boundary, and rhetorical plan.
For formal academic or professional prose, use academic-self-review-loop after cognitive-frameworks and before style/document-quality gates. The loop should identify concrete weaknesses, revise argument quality, and make a fresh second-pass judgement.
When language quality, British English, or the user's preferred output pattern matters, use uk-academic-writing-style and style-memory-and-revision-gate before delivery.
When the user asks for AI-looking prose cleanup, de-AI, humanising, lower-AI-rate, AIGC, AI detector, or disclosure-hiding help, use authorial-voice-integrity to reframe the task as authorial voice, academic/professional integrity, and evidence-led style. For formal artifacts, source-first, Material Passport, and integrity checks still come before substantive style revision. Do not promise detector outcomes or use evasion tactics.
Use this quick sequence at the start of every Production Window turn:
research-wiki/STAGE_GRAPH.md, write a Stage Continuity Capsule, and run scripts/stage_continuity_capsule_check.py when available.SKILL.md files before acting on any non-trivial task.research-wiki/PRODUCTION_RUN_REGISTER.md for substantial Production tasks.Use this receipt wording:
Skill routing:
- Mode:
- Skills selected:
- Why these skills:
- Skills considered but skipped:
- Subagent decision:
- Gates required:
- Gates completed:
- Recall Decision Note:
- Stage Continuity Gate triggered:
If a substantial Production output has no skill-routing receipt, treat the task as incomplete for maintenance-audit purposes.
Choose one primary mode before choosing skills:
| Mode | Use When | Main Rule |
|---|---|---|
| Research Mode | literature search, websites, LMS/portals, supervisor/PI/client or public resources, policy, source discovery | use dissertation-research-search-protocol; use dissertation-learning-loop after useful new reading; separate confirmed source information from interpretation |
| Review Mode | checking proposal, ethics, interview guide, source notes, citations, risks | findings and risks first; use source paths and evidence boundaries |
| Drafting Mode | writing proposals, manuscripts, reports, protocols, stakeholder notes, ethics/compliance wording, Word files | use source-first gate before drafting; use relevant style, style memory, project-delivery, and document-quality gates before delivery |
| Integration Mode | updating project rules, skills, Obsidian, task state, knowledge base | preserve existing project conventions and record what changed |
| Maintenance Mode | false-run checks, system reliability, skill stocktake, workspace surface, automations, knowledge ops | audit first, fix confirmed defects only, and update task state |
Classify the user's request into one or more categories:
| Task Type | Primary Skills |
|---|---|
| project setup / project type adaptation | research-project-adapter, agent-orchestration, dissertation-knowledge-ops |
| project memory / notes / supervisor discussion | dissertation-knowledge-ops, dissertation-learning-loop, supervisor-feedback-loop, context-continuity |
| proposal / manuscript / report / grant writing | research-project-adapter, dissertation-source-first-gate, material-passport, academic-integrity-preflight, cognitive-frameworks, dissertation-argument-spine, dissertation-chapter-plan, dissertation-research-review, academic-self-review-loop, authorial-voice-integrity, uk-academic-writing-style, style-memory-and-revision-gate, dissertation-document-quality-gate |
| ethics / compliance / participant materials | dissertation-source-first-gate, responsible-ai-agent-audit, dissertation-research-review, dissertation-shared, dissertation-document-quality-gate |
| LMS/module requirements | dissertation-research-search-protocol, dissertation-knowledge-ops, dissertation-chapter-plan, dissertation-citation-audit |
| literature search / literature review | dissertation-research-search-protocol, dissertation-learning-loop, dissertation-literature-review, cognitive-frameworks, dissertation-argument-spine, dissertation-citation-audit, academic-self-review-loop when drafting formal synthesis |
| bounded source planning / reading priority sorting | dissertation-source-first-gate, dissertation-research-search-protocol, dissertation-citation-audit, dissertation-learning-loop, context-continuity; do not add the full formal-writing chain unless the user asks for formal prose, Word/DOCX, stakeholder-facing output, submission-facing output, or protected source-of-record edits |
| bounded research lookup | dissertation-research-search-protocol, dissertation-citation-audit, context-continuity; label metadata-only, source-section-reviewed, or evidence-insufficient status and do not update source registers without confirmation |
| minor edit / typo / citation-key formatting | dissertation-source-first-gate, dissertation-citation-audit, context-continuity; preserve the no-substantive-change boundary and do not upgrade citation readiness |
| research questions / methodology | cognitive-frameworks, dissertation-argument-spine, dissertation-research-review, dissertation-chapter-plan, academic-self-review-loop when drafting formal prose |
| Stage Continuity triggered deliverable | context-continuity, cognitive-frameworks, dissertation-source-first-gate, dissertation-argument-spine, dissertation-research-review |
| interview guide / data collection | qualitative-theme-audit, responsible-ai-agent-audit, dissertation-research-review |
| confirmed design-elicitation / co-design outputs | qualitative-theme-audit, responsible-ai-agent-audit, dissertation-research-review; restore archived co-design/design-spec skills only after Maintenance confirms a recurring dedicated deliverable |
| AI agent concept / prototype | cognitive-frameworks, responsible-ai-agent-audit, dissertation-research-review; restore archived design/prototype skills only for a dedicated specification or prototype audit |
| adoption conditions | responsible-ai-agent-audit, cognitive-frameworks, dissertation-research-review |
| active learning or intervention design | cognitive-frameworks, responsible-ai-agent-audit, dissertation-research-review; restore archived topic packs only when this is a concrete project phase |
| teaching knowledge base / RAG plan | dissertation-knowledge-ops, dissertation-research-search-protocol, context-continuity |
| unclear research route / high-impact idea / early design discussion | brainstorming, dissertation-argument-spine, dissertation-research-review |
| file-to-Markdown conversion / source ingestion from documents | markitdown, dissertation-source-first-gate, dissertation-knowledge-ops, dissertation-learning-loop |
| learning loop / source ingestion | dissertation-learning-loop, dissertation-knowledge-ops, context-continuity |
| academic writing / authorial voice / AI-style cleanup | authorial-voice-integrity, academic-integrity-preflight when disclosure or prompt-residue risk appears, academic-self-review-loop, uk-academic-writing-style, style-memory-and-revision-gate, dissertation-document-quality-gate |
| rubric / marking criteria / journal / funder / client requirement audit | research-project-adapter, dissertation-source-first-gate, dissertation-research-search-protocol, dissertation-argument-spine, dissertation-research-review, dissertation-citation-audit, dissertation-chapter-plan |
| LMS / portal / requirement / deadline / word count / submission rule | research-project-adapter, dissertation-source-first-gate, dissertation-research-search-protocol, dissertation-knowledge-ops, dissertation-document-quality-gate |
| major chapter rewrite / argument audit | dissertation-argument-spine, dissertation-research-review, dissertation-citation-audit |
| final review / defense / viva | dissertation-research-review, context-continuity; restore archived viva-prep only when a thesis/dissertation defense phase is active |
| long task / context handoff / compaction risk | context-continuity, dissertation-knowledge-ops |
| formal document delivery / QA | cognitive-frameworks when argument planning is needed, academic-self-review-loop, uk-academic-writing-style, style-memory-and-revision-gate, dissertation-document-quality-gate, context-continuity |
| agent self-debug / false-run recovery | dissertation-agent-self-debug, dissertation-workspace-surface-audit, context-continuity |
| GitHub release / public template visible-surface verification | release-surface-verification, dissertation-agent-self-debug, dissertation-workspace-surface-audit when live browser or rendered-surface checks are needed |
| agent architecture audit / rule conflict | dissertation-agent-architecture-audit, dissertation-skill-stocktake, context-continuity |
| workspace surface audit | dissertation-workspace-surface-audit, dissertation-automation-audit, dissertation-knowledge-ops |
| browser automation / LMS browser inspection / local page verification | playwright-dissertation-browser, global playwright, dissertation-research-search-protocol when source capture is involved |
| automation / hook / connector audit | dissertation-automation-audit, dissertation-workspace-surface-audit, context-continuity |
| skill stocktake / skill quality review | dissertation-skill-stocktake, dissertation-agent-architecture-audit, context-continuity |
| knowledge system maintenance | dissertation-knowledge-ops, dissertation-workspace-surface-audit, context-continuity |
| agent system maintenance / bug check | dissertation-agent-self-debug, dissertation-agent-architecture-audit, dissertation-skill-stocktake, dissertation-workspace-surface-audit, context-continuity |
| project skill/rule migration | using-superpowers, project-skill-creator-governance, global skill-creator, dissertation-skill-stocktake, dissertation-agent-architecture-audit, dissertation-document-quality-gate, context-continuity |
| Superpowers-style workflow request | using-superpowers, brainstorming when idea/design is unclear, then the relevant dissertation or maintenance skills |
When creating or updating Codex skill files, use the system skill-creator skill if it is available. For external workflow migration, audit first, then adapt into existing dissertation skills unless the user explicitly asks for a new active skill.
For project skill creation or updates, use project-skill-creator-governance before global skill-creator. Do not copy global skill-creator or global playwright into .agents/skills/ unless the user explicitly asks for frozen local copies.
Archived adapter or topic-pack skills under .agents/archived-skills/ are outside the active discovery root and are not routing targets. If a task appears to need one, first use the active core route and only move the required pack back into .agents/skills/ after Maintenance confirms a recurring project-phase need that justifies the added context load.
For Superpowers-style workflows, use using-superpowers as an adapter. It must not override source-first, rubric evidence, document-quality, privacy, or window-separation rules.
For file-to-Markdown conversion, use markitdown only after checking tool availability and privacy boundaries. If MarkItDown is unavailable, do not install it without explicit confirmation.
Use one skill when:
Use multiple skills when:
For formal documents, apply this gate before writing:
TO CONFIRM fields.Use the dedicated dissertation-source-first-gate skill for full instructions when the task involves formal documents, personal/admin facts, official requirements, participant-facing materials, citations, or stakeholder-facing drafts.
For rubric, grade-band, marking criteria, LMS/portal requirements, journal/funder/client requirements, deadlines, word counts, or submission rules:
university-guidance/RUBRIC_EVIDENCE_GATE.md when assessed academic grading is relevantFor school templates with colored guidance:
xxx or blue guidance only with source-grounded or clearly marked draft wordingConsider subagents under the user's standing permission, but apply the safety and usefulness tests below.
Use subagents when:
Avoid subagents when:
For broad research:
For LMS/module requirement work:
For proposal/ethics preparation:
For AI agent design:
TO CONFIRM.dissertation-document-quality-gate before delivering important Word files or formal project outputs.research-wiki/DOCUMENT_PIPELINE.md for important Word, PDF, or stakeholder-facing outputs so thinking, writing, and delivery checkpoints are recorded or explicitly marked not applicable.For complex tasks, briefly state:
For simple tasks, do not over-explain; just use the relevant skill and proceed.
For substantial Production tasks, the final delivery or run receipt must also include: