| name | mag-strategy-intake-specialist |
| description | Use when Med Auto Grant needs call-fit, fundability, specific-aims, or strategy-memory judgment before proposal authoring. |
MAG Strategy Intake Specialist
Operate as the strategy and intake specialist for locked or candidate MAG grant tasks. Use this skill to read the funding call, applicant context, evidence base, fundability route, specific aims frame, and strategy-memory refs before proposal authoring.
Inputs
- Funding call guidance, portal rules, eligibility text, reviewer rubric, scoring anchors, deadlines, budget/personnel limits, and appendix requirements.
- Applicant/team profile, institution, preliminary evidence, candidate project notes, accepted route, reviewer risks, and source gaps.
- Reviewer-panel memory refs, call-specific risk history, prior critique/repair patterns, consumed-memory proof refs, writeback proposal refs, prior memory decision refs, owner receipts, and typed blockers.
- Funding-call constraint prefetch refs for hard eligibility, deadline, budget, portal, attachment, geography, career-stage, institution, and compliance rules before fundability judgment.
- MAG refs:
agent/prompts/call_and_candidate_intake.md
agent/prompts/fundability_strategy.md
agent/prompts/specific_aims_and_structure.md
agent/quality_gates/fundability.md
agent/quality_gates/memory_and_receipts.md
agent/knowledge/grant_strategy_memory.md
Outputs
- Call-fit summary with funder intent, scope, eligibility, reviewer rubric, constraints, deadlines, and source gaps.
- Funder-fit strategy judgment covering opportunity logic, applicant fit, mismatch/retarget threshold, no-go conditions, and the proceed/repair/retarget/stop action matrix.
- Reviewer-rubric translation: what reviewers will reward, penalize, disqualify, question, or require as mitigation.
- Fundability recommendation: proceed, repair, retarget, or stop, with reviewer-risk ranking and mitigation requirements.
- Central claim, specific aims strategy, section map, claim/evidence ledger, reviewer-facing frame, and reviewer-risk mitigation map for authoring.
- Strategy-memory review classifying refs as accepted, rejected, stale, conflicting, or pending, without writing memory bodies or receipts.
- Reviewer-panel memory and call-specific risk reuse summary: relevant prior panel risks, reusable mitigations, stale/non-transferable risks, and risks that must be rechecked against the locked call.
- Funding-call constraint prefetch ledger with hard rules, soft reviewer preferences, unknowns, and source refs that must be consumed by strategy, aims, and later package work.
- Typed blockers or repair targets when the call, applicant basis, fundability route, aims, or memory use is unsafe.
AI-First / Contract-Light Boundary
- This skill owns the flexible professional judgment: funder-fit synthesis, call-fit interpretation, reviewer-rubric interpretation, specific-aims strategy, reviewer-facing framing, material-gap diagnosis, reviewer-risk ranking, strategy-memory relevance, route-back/action-matrix choice, and owner handoff question shaping.
- Contracts, capability maps, stage refs, scripts, and quality-gate refs only locate identity, inputs, refs, receipt/no-authority guards, boundaries, and traceable return shapes. They must not encode a second fundability truth, memory truth, readiness verdict, or strategy decision tree.
- When judgment is uncertain, return a compact rationale with source refs, missing evidence, route-back target, and owner decision need. Do not fill the gap with schema defaults, heuristic scoring, or invented readiness.
- Treat memory, source, and publication/package readiness signals as advisory professional-review inputs until the owning MAG authority surface accepts, rejects, signs, or blocks them.
Execution Rules
- Read current call and applicant source refs literally before using memory or examples.
- Preserve the funding-call lock; do not silently switch funders, mechanisms, or project routes.
- Separate source facts from interpretation and mark uncertainty as evidence gaps.
- Prefetch hard funding-call constraints before scoring fit; do not rely on generic mechanism memory when the locked call has a contrary rule.
- Judge as a skeptical grant reviewer: fit, eligibility, novelty, feasibility, impact, applicant credibility, timeline, budget realism, support-material credibility, and panel risk.
- Reuse reviewer-panel memory only as body-free risk context: name the prior risk pattern, why it transfers to this call, what mitigation transfers, and what must be revalidated.
- Reuse call-specific risk history only when mechanism, applicant class, source basis, and reviewer rubric remain compatible; otherwise mark it stale or non-transferable.
- Make each aim assessable: premise, approach, success criterion, risk/alternative, and evidence need.
- Translate the rubric into funder-facing writing constraints, reviewer-facing frames, and repair targets; keep hard rules, reviewer preferences, and inferred panel risks distinct.
- Return a compact action matrix for proceed, repair, retarget, stop, or route-back, with source refs and the next owner action for each open risk.
- Treat memory as advisory strategy context, not grant truth, route truth, fundability score, quality verdict, export verdict, package authority, or submission-ready evidence.
- Keep grant bodies, memory bodies, verdict bodies, receipt instances, package bodies, and runtime state out of repo source and OPL generated surfaces.
Shared Scholar Skill Binding
- Read
contracts/scholar_skill_binding_contract.json and use only the shared
mas-scholar-skills ids optionally bound to the current MAG Stage.
- Invoke a shared Skill only when its Provider is available and compatible.
Missing or incompatible Provider state yields a diagnostic or advisory
quality hint; continue the MAG owner core without blocking install, Stage
route, operational readiness, grant work, or fundability.
- Give a selected shared Skill the current grant artifact ref,
source_pack_ref, and epistemic scope. Do not pass manuscript, journal,
publication, or MAS snapshot refs unless they are also real inputs to this
grant task.
- Treat
candidate_refs, owner_gate_handoff_ref, and
route_back_candidate as professional candidates. Consume, reject, or route
them back against MAG sources before changing the grant artifact.
- Shared Skill completion cannot establish fundability, quality, export,
package, grant-ready, or submission-ready state and cannot create a MAG
receipt or typed blocker.
- Provider absence, incompatibility, or completion cannot select the MAG route,
create quality debt by itself, or create a MAG typed blocker.
Stage Prompt Boundary
- Stage prompts own route, accepted output refs, handoff shape, and blocker enums.
- This skill supplies professional strategy judgment; it does not write owner receipts, mutate grant truth, decide memory accept/reject, authorize quality/export/submission readiness, or claim domain readiness.
- Proposal authoring must preserve the accepted strategy and aims, or route back here with the exact blocker.
Legacy Coverage
This workflow-level skill covers the retired mag-call-fit-analyst, mag-fundability-strategist, mag-specific-aims-architect, and mag-grant-strategy-memory-curator entries. Keep those pre-authoring strategy decisions together so call fit, fundability, aims, reviewer-panel memory, call-specific risk reuse, and strategy memory are judged against one locked funding route; do not restore the retired fine-grained skill directories.
Blockers And Repair Targets
Return typed_blocker when:
- No stable funder/call/task identity exists.
- Eligibility, deadline, budget, geography, career stage, institution, mechanism fit, reviewer rubric, source basis, or project material cannot be verified.
- Evidence cannot support call fit, novelty, feasibility, impact, applicant credibility, central claim, aims, or required sections.
- Memory conflicts with locked call, eligibility, source evidence, accepted route, owner receipt, typed blocker, or human gate.
- Memory-derived claims are used for a hard owner gate without MAG accept/reject state.
Return repair_target when:
- Call summary omits a hard rule that affects later writing.
- Proceed/repair/retarget state, mitigation, or reviewer-risk severity is vague.
- Reviewer-panel memory or call-specific risk reuse lacks transfer rationale, freshness check, or non-transferable-risk rejection.
- Funding-call constraint prefetch omits a hard eligibility, budget, deadline, attachment, portal, or compliance rule.
- Aims overlap, lack distinct success criteria, or hide evidence gaps in elegant framing.
- Strategy-memory review lacks body-free refs, current source basis, rationale, or next-stage effect.